You are on page 1of 21

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/258078060

Shear thinning effects on blood flow in straight and curved tubes

Article  in  Physics of Fluids · July 2013


DOI: 10.1063/1.4816369

CITATIONS READS
30 2,169

2 authors, including:

Erica M Cherry Kemmerling


Tufts University
18 PUBLICATIONS   356 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

BIM energy View project

All content following this page was uploaded by Erica M Cherry Kemmerling on 03 April 2015.

The user has requested enhancement of the downloaded file.


Shear thinning effects on blood flow in straight and curved tubes
Erica M. Cherry and John K. Eaton

Citation: Physics of Fluids (1994-present) 25, 073104 (2013); doi: 10.1063/1.4816369


View online: http://dx.doi.org/10.1063/1.4816369
View Table of Contents: http://scitation.aip.org/content/aip/journal/pof2/25/7?ver=pdfcov
Published by the AIP Publishing

Articles you may be interested in


A file of red blood cells in tube flow: A three-dimensional numerical study
J. Appl. Phys. 116, 124703 (2014); 10.1063/1.4896358

Modeling of the blood rheology in steady-state shear flows


J. Rheol. 58, 607 (2014); 10.1122/1.4866296

Orbital drift of capsules and red blood cells in shear flow


Phys. Fluids 25, 091902 (2013); 10.1063/1.4820472

Chaotic flow in an aortic aneurysm


J. Appl. Phys. 113, 214909 (2013); 10.1063/1.4809559

Reducing the data: Analysis of the role of vascular geometry on blood flow patterns in curved vessels
Phys. Fluids 24, 031902 (2012); 10.1063/1.3694526

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
PHYSICS OF FLUIDS 25, 073104 (2013)

Shear thinning effects on blood flow in straight


and curved tubes
Erica M. Cherrya) and John K. Eaton
Mechanical Engineering, Stanford University, Building 500, Room 500A,
488 Escondido Mall, Stanford, California 94305, USA
(Received 20 December 2012; accepted 8 July 2013; published online 26 July 2013)

Simulations were performed to determine the magnitude and types of errors one can
expect when approximating blood in large arteries as a Newtonian fluid, particu-
larly in the presence of secondary flows. This was accomplished by running steady
simulations of blood flow in straight and curved tubes using both Newtonian and
shear-thinning viscosity models. In the shear-thinning simulations, the viscosity was
modeled as a shear rate-dependent function fit to experimental data. Simulations in
straight tubes were modeled after physiologically relevant arterial flows, and flow
parameters for the curved tube simulations were chosen to examine a variety of sec-
ondary flow strengths. The diameters ranged from 1 mm to 10 mm and the Reynolds
numbers from 24 to 1500. Pressure and velocity data are reported for all simula-
tions. In the straight tube simulations, the shear-thinning flows had flattened velocity
profiles and higher pressure gradients compared to the Newtonian simulations. In
the curved tube flows, the shear-thinning simulations tended to have blunted axial
velocity profiles, decreased secondary flow strengths, and decreased axial vorticity
compared to the Newtonian simulations. The cross-sectionally averaged pressure
drops in the curved tubes were higher in the shear-thinning flows at low Reynolds
number but lower at high Reynolds number. The maximum deviation in secondary
flow magnitude averaged over the cross sectional area was 19% of the maximum sec-
ondary flow and the maximum deviation in axial vorticity was 25% of the maximum
vorticity. 
C 2013 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4816369]

I. BACKGROUND
It is widely known that blood is a complex fluid that contains deformable cells. However, many
blood simulations have approximated it as Newtonian.1–6 Some of these papers do not mention
the non-Newtonian nature of blood and others explain their constant viscosity model by citing
McDonald’s statement that blood flow in large arteries is approximately Newtonian.7 In reality,
blood exhibits both shear-thinning and viscoelastic behavior, and this study focuses on shear-
thinning. Although it is true that blood’s viscosity becomes approximately constant at high shear
rate, large blood vessels contain a region of low shear rate near the center where the viscosity could
be an order of magnitude or more larger than the viscosity near the vessel walls. In this paper, we
will examine the statement that shear-thinning behavior of blood can be ignored in large arteries and
evaluate the regimes where this statement is valid.
It is important to recognize that shear-thinning is not the only non-Newtonian characteristic
of blood. Blood also has viscoelasticity, which is caused primarily by the coagulation of columns
of red blood cells to form coherent elastic structures.8–12 The membranes of individual red blood
cells have also been shown to be elastic.13–15 However, studies have shown that blood viscoelasticity

a) Author to whom correspondence should be addressed. Electronic mail: echerry@stanford.edu

1070-6631/2013/25(7)/073104/19/$30.00 25, 073104-1 


C 2013 AIP Publishing LLC

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-2 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

is only detectable in unsteady flow16, 17 and even then is only significant at shear rates less than
10 s−1 .18, 19 Because the flows considered in this study are steady and contain only very small regions
with shear rates lower than 10 s−1 , the blood viscosity model does not include viscoelasticity. Further
justification for this is provided by Bodnar et al.20 who simulated both steady pure shear-thinning
and combined viscoelastic and shear-thinning flow with a geometry and Reynolds number similar
to those in the current study. They found negligible difference between the two flows.
There have been a large number of previous studies on blood rheology. A thorough review of
the published work is given by Pries et al.,21 and additionally, Alonso et al.22 published a study on
hysteresis effects on blood viscosity in unsteady flow. Most of the papers reviewed by Pries et al.21
report measurements from rotational viscometers with gaps less than 1 mm and pipettes between 3
and 500 μm in diameter. However, none of these studies were for conditions close to those in a large
artery.
Several different approaches have been taken in simulating the non-Newtonian nature of blood.
One common method is to model arterial blood flow as a two-phase material with a core region rich in
red blood cells and a Newtonian pure plasma boundary layer.23–25 Others have modeled the boundary
layer as a region of reduced but nonzero hematocrit.26 Two-phase models have the advantage of accu-
rately capturing the Fahraeus-Lindqvist effect, in which red blood cells preferentially concentrate at
the center of vessels and leave an evacuated plasma layer near the walls. However, piecewise-defined
viscosity distributions tend to produce velocity profiles with discontinuous derivatives at the phase
boundaries.
Another common technique is to use a constitutive model for blood viscosity based on statistical
descriptions of erythrocyte motion and collisions. Such studies generally use a simplified geometric
model for an erythrocyte such as a sphere,23 an ellipse,27 or a dumbell,28 although in recent years, an
increasing number of studies have used more realistic models.29 These models have the advantage
of accounting for the particulate nature of blood and may be more accurate in capillary tubes, where
the presence of a single cell can significantly perturb the flow field. However, models of this type
require assumptions about the shape and orientation of red blood cells and how they interact during
collisions. It is difficult to evaluate the accuracy of these assumptions. Statistical models are also
computationally intensive, which makes them unrealistic for flows in large arteries.
The last category of common blood viscosity models is continuous viscosity functions, which
assign a viscosity to every point in the simulation domain based on the local shear rate, position,
or time history. Some of these models are strictly shear-dependent30–32 and others account for
the viscoelasticity.33, 34 Continuous viscosity models have relatively low computational cost and
produce smooth velocity profiles with continuous velocity gradients. However, it is difficult to assess
the regimes in which these models are accurate because although many studies have measured the
viscosity of blood in micron to millimeter-sized tubes and channels, there is very little experimental
work on blood viscosity in tubes the size of medium and large arteries.21 The model used in this
study is of the continuous viscosity function type and uses a fit to the data of Brooks et al.35 This
dataset was chosen for its low uncertainty and wide range of shear rates.
Although few datasets on the quantitative effect of blood’s shear thinning behavior in large
arteries are available, several papers provide experimental evidence that blood does not behave like
a Newtonian fluid in large arteries. Shahada et al.,36 Osada and Radegran,37 Kiserud et al.,38 and
Ade et al.39 all collected in vivo blood velocity data and found velocity profiles that were blunted
compared to profile shapes for Newtonian flow. Most other experimental studies on blood flow in
large arteries report only flow rate1, 40, 41 or velocity data with resolution too low to adequately judge
whether the flow is Newtonian.3, 42–44 There have also been a number of experimental studies of
velocity profiles in microcirculation.45–48 These studies all found blunter-than-parabolic velocity
profiles.
Nearly all of the computational and in vitro experimental studies mentioned above presented
data on blood flowing in a straight tube or channel. Agrawal et al.49 studied the flow of power-law
non-Newtonian fluids in curved tubes, but their viscosity distributions were all shear-thickening. The
only study of blood flow in a curved tube the authors could find was Gijsen et al.,50 who examined
non-Newtonian unsteady flow in a curved tube and found significant variation in velocity profiles
between Newtonian and non-Newtonian cases.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-3 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

Most of the studies above, such as El-Khattib et al.,32 have looked at the effect of shear thinning in
parallel flows, which have only a single velocity component. These studies have provided convincing
evidence that in steady and unsteady straight tube flows, shear-thinning behavior significantly alters
velocity profiles and pressure gradients where mass flow rate is fixed. Real human arteries are
tortuous and contain secondary flows, defined as velocity components that are not parallel to the
local tube axis. Thus, an important next step is to determine whether secondary flows change the
importance of shear-dependent viscosity variation.
In this paper, we address this question by studying shear-thinning flow in curved tubes, where
Newtonian fluids are known to develop strong secondary flows. The only previous work mentioning
secondary flows in shear thinning fluids was by Gijsen et al.,50 and because their flow was unsteady,
it is difficult to separate the influence of time-dependence from that of secondary flow. In this paper,
we isolate the effect of secondary flows by simulating steady flows with a range of secondary flow
strengths.

II. FLOW GEOMETRIES AND METHODS


Table I summarizes the tube geometry and flow parameters for all of the simulations. The first
five cases are fully developed laminar flow in straight round tubes. The tube diameters were chosen
to be representative of the flow in various arteries as designated by the case names. For each case,
the mass flow rate was based on averaged published arterial flow rate data.51–56 Thus the shear rates
cover the range encountered in realistic artery flows. We ran two different simulations for each tube
diameter. The first used the shear thinning model discussed in Sec. III, while the second treated the
flow as Newtonian. For the Newtonian simulations, a dynamic viscosity of 0.004 kg/m s was used
because this is the asymptotic, high shear rate viscosity from the shear-thinning model.
Fully developed straight tube flows are parallel, so the radial and tangential velocity components
are identically zero. Real blood vessels are not straight, nor are they perfectly round. Therefore, they
are likely to have significant secondary flows. Secondary flows produce complicated distributions of
shear rate, which may produce unexpected effects of shear thinning. Secondary flows also produce
axial vorticity which is not present in parallel flows. To investigate these flow features, two tubes
with circular arc curves were simulated. The radius of curvature and the Reynolds number were
varied independently since they both affect the strength of secondary flows in curved tubes. Both
curved tubes had the same arclength (3.77 cm) but one had a radius of curvature three times that
of the other. Both curved tubes had 8 mm diameters to match the experimental data of Bovendeerd
et al.57 for validation. Both geometries contained two-diameter-long straight sections upstream of
the inlet and downstream of the outlet of the curved sections. Eight curved-tube simulations were
run to cover every possible combination of Newtonian or shear-thinning fluid, high or low curvature,
and high or low Reynolds number. The flow parameters for these simulations are also summarized in
Table I. In shear-thinning flows, the Reynolds number is ill-defined because the viscosity is a function

TABLE I. Simulation summary. Reynolds numbers are calculated using the bulk velocity, the tube diameter, and the
Newtonian or asymptotic high shear rate viscosity (where applicable). All simulations were performed for both Newtonian
and shear-thinning fluid.

Case Grid Diameter (cm) Bulk velocity (cm/s) Re

Straight femoral Straight 1.0 7 185


Straight subclavian Straight 0.8 6 127
Straight splenic Straight 0.5 15 199
Straight hepatic Straight 0.3 28 223
Straight intercostal Straight 0.1 9 24
90◦ high Re 90◦ 0.8 70 1500
90◦ low Re 90◦ 0.8 7 150
30◦ high Re 30◦ 0.8 70 1500
30◦ low Re 30◦ 0.8 7 150

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-4 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

FIG. 1. Shear rate vs. blood viscosity data35, 60–63 and fit.

of space if the shear is nonconstant. In this paper, all Reynolds numbers given for shear-thinning
flows are based on the asymptotic high shear rate blood viscosity (μ = 0.004 kg/ms).

III. SHEAR THINNING MODEL


The viscosity was modeled as a function of local shear rate, where local shear rate was defined
as the 2-norm of the local velocity gradient tensor. The function was a fit to the data of Brooks et al.35
for a hematocrit of 42%, a reasonable value for large arteries in a healthy adult. Brooks et al.35 did
not measure viscosities at this hematocrit, so the fit was based on an average of their data for 35.9%
and 48.0% hematocrit. The fit is given by Eq. (1), where μ is the kinematic viscosity and s is the
natural log of the local shear rate
 
1 −0.00795s + 0.344 −0.0128s + 0.398
μ= + 2 . (1)
2 s 2 + 9.02s + 25.0 s + 6.11s + 14.5
Figure 1 shows the fit and the data, along with some other experimental datasets. It is clear from
this plot that the data of Brooks et al.35 are well supported by other studies and cover a larger range
of shear rates than most. The sensitivity of the flow to this particular choice of viscosity model
will be examined in Sec. VI. No measurements were available for shear rates below 0.17 s−1 , so
μ = 0.046 kg/ms (the value of the viscosity fit at 0.17 s−1 ) was taken as a constant for shear rates
smaller than 0.17 s−1 to avoid a viscosity singularity at zero shear. The viscosity was also taken as a
constant of μ = 0.004 kg/ms above shear rates of 700 s−1 , where there were no experimental data,
because the data for lower shear rates were approaching a clear asymptote near this point.

IV. GRIDS AND NUMERICAL METHOD


The straight and curved tube grids were generated in Gambit and summaries of their geometries
are given in Table II. A sample grid is shown in Figure 2. The mesh on the inlet face was created
first and was identical in all three grids. First, a growth layer ten cells deep with 40 cells in the
circumferential direction was applied to the face boundary. The thinnest cell (0.12 mm thick) was
at the edge and the inner cells had a growth factor of 1.1 in the radial direction. The remaining face
area was gridded using Gambit’s quadrilateral pave function and was smoothed using Laplacian
smoothing. The volumes were gridded by sweeping the face grid along the axis of the straight tube
and the outer radius of the curved tubes. The straight tube grid elements were 0.1 tube diameter long
in the axial direction and the curved tube grid elements were 0.1 tube diameter long at the widest
part of the elbow. The curved tube elements were shorter near the inside of the elbow in order to
preserve the shape of the grid in any given flow-normal cross section.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-5 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

TABLE II. Grids.

Number of Element Radius of Turning angle


Grid elements shape curvature (deg) Dimensions

Straight 31 320 Hexahedral ∞ 0 d = 0.1, 0.3, 0.5, 0.7, 1.0,


2.5 cm
30◦ 61 264 Hexahedral 7.2 cm 30◦ d = 0.8 cm, arclength
= 3.77 cm, upstream duct
length 2 × d
90◦ 61 264 Hexahedral 2.4 cm 90◦ d = 0.8 cm, arclength
= 3.77 cm, upstream duct
length 2 × d

For the straight tube cases, the boundary conditions were periodic and the mass flow rate was
specified. For the curved tube cases, the inlet velocity profile shape was specified as a parabola in
Newtonian flows and the fully developed shear-thinning velocity profile in shear-thinning flows.
These profile shapes were multiplied by a constant to produce the bulk velocities listed in Table I.
The outlets of the curved tube simulations were set at uniform, atmospheric pressure.
All simulations were performed in the commercial software Fluent using a second order upwind
solver. The details of this numerical method are explained in Ref. 58. The shear-thinning model was
implemented by using a user-defined function to assign a custom viscosity to each grid cell. At each
solver iteration, the local shear rate was computed in each cell and a corresponding viscosity was
assigned at that location to be used in the next iteration. Simulations were considered converged
when the residuals had decreased by eight orders of magnitude relative to the initial guess. The
implementation of the shear-thinning model caused an average increase in computational cost of
4.6% compared to the Newtonian model. This number was calculated by comparing the time taken
to compute one solution iteration for otherwise identical Newtonian and shear-thinning simulations.
However, in some cases, the shear-thinning cases took more iterations to converge.

V. SIMULATION VALIDATION
The simulation was validated in two stages. The first confirmed that the simulation reproduced
known solutions for Newtonian flow and the second proved that it correctly solved for the velocity
profile of a shear-thinning fluid by comparison to an analytical flow solution. For the Newtonian
solution validation in the straight tube, fully developed flow of several different constant viscosities
was simulated and compared with the Poiseuille solution. All velocity profiles matched the analytical
solution within 0.1% and all pressure gradients matched within 0.3%.
Validating the curved tube simulations was somewhat more complicated. There is no analytical
solution for flow in a curved tube, so comparison to experimental results was necessary. A validation

FIG. 2. Sample grid, side view (left), and end-on view (right).

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-6 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

FIG. 3. Comparison of the experimental data of Bovendeerd et al. (solid) and computed (dashed) streamwise velocity profiles
at various angles from the inlet of a 90◦ elbow. The diagram at the top shows the plane and direction of the velocity profile
extraction.

simulation was run in the 90◦ elbow geometry replicating the flow conditions of Bovendeerd
et al.,57 who took laser Doppler velocity measurements on flow in an identical geometry. The
Reynolds number was 700 and the bulk velocity was 1.09 m/s. The flow of Bovendeerd et al.57
was fully developed upstream of the elbow, so the simulation’s inlet velocity profile was a parabola.
Bovendeerd et al.57 took streamwise and in-plane velocity data at seven streamwise positions along
the curved portion of their tube. Comparisons of streamwise and in-plane velocities between the
experiment and simulation are shown in Figures 3 and 4.
Streamwise velocity profiles were taken across the tube pointing towards the center of the elbow,
as shown in Figure 3. The coordinates are taken from Bovendeerd et al.,57 with the nondimensional
radius defined as
Relbow − r
R= , (2)
Rtube
where r is the outward radial distance from the center of the elbow.
At the start of the elbow, the axial velocity is barely distorted from a parabola and the simulation
and experiment agree very well. Downstream, the apex of the profile gradually shifts towards the
center of the elbow and then develops an inflection point around the 23.4◦ profile. At this point, the
computed and experimental data are no longer perfectly aligned, but their agreement is always within
5% of the average velocity. Moving downstream, the velocity profile becomes highly distorted with
a peak near the inner radius.
The qualitative features and both the location and magnitude of the peak velocity are in excellent
agreement between simulations and experiment. However, the computed profiles tend to have sharper
features than the experimental profiles, which may be due to the filtering effect Bovendeerd et al.57
cite as a result of the finite size of their laser probe. The size of this probe could have caused
velocity averaging over lengths up to 5.8% of the diameter. The most prominent deviation between
the experimental and computed profiles occurs near the center of the tube in the 39.8◦ profile, where
the computed velocity is less than the experimental profile by about 20% of the bulk velocity. The
final velocity profile at 81.9◦ shows a highly distorted velocity profile with the peak velocity about

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-7 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

FIG. 4. Comparison of the experimental data of Bovendeerd et al. (top) and computed (bottom) in-plane velocities at various
angles from the inlet of a 90◦ elbow.

1/7 a diameter away from the inner wall. The computed and experimental data agree within 10%
throughout this profile and the peak level and the velocity in the plateau in the center of the tube are
almost identical.
Figure 4 compares experimental and computed secondary flows. The figure was generated
by copying figures from Bovendeerd et al.57 for the top half of each cross section and plotting
computed secondary flows on the bottom half. The computed vectors were scaled to match the size
of the figures of Bovendeerd et al.57 The 0◦ plane of data is not plotted because the secondary flows
were negligible there and the 4.6◦ plane shows only weak secondary flows. However, by the 11.7◦
plane, a vortex pair with peak velocities around 20% of the bulk velocity has formed, and these
velocities have increased to about 25% of the bulk velocity by the 23.4◦ plane. At this point, the
vortex pair starts to migrate towards the outside of the elbow. By the 58.5◦ plane, its center is offset
from the centerline by about 14% of the diameter. At the 81.9◦ plane, the vortex pair is still present
but has weakened considerably. The simulation captures the shape of the secondary flows well
throughout the elbow. The magnitude agreement is good when the flow is fast (11.7◦ plane through
58.5◦ plane) but the simulation slightly over-predicts the speed of the secondary flows when they are
slower.
The second stage of simulation validation involved comparing simulation data to an analytical
solution for shear-thinning flow in a round tube. The analytical solution was derived by assuming
the following velocity profile as the axial velocity solution for fully developed flow in a round tube
for a shear-thinning fluid:
  r 8 
u = u max 1 − . (3)
R

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-8 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

FIG. 5. Comparison of computed and exact velocity profiles for a shear-thinning fluid with varying values of threshold shear
rate.

This profile was chosen because it is flattened relative to the parabolic profile in Newtonian flow
and the simple algebraic form allows for closed form analysis. Because a fully developed pipe flow
must obey the linear shear law (Eq. (4)), it is possible to solve for the viscosity by plugging in the
expression for the assumed velocity profile:
∂u
τ =μ ∼ r, (4)
∂r
 −6/7
−6 ∂u
μ = C1r = C2 , (5)
∂r
where C1 and C2 are constants. If the algebraic shear thinning law (Eq. (5)) is used in the momentum
equation, the resulting solution should match the assumed velocity profile (Eq. (3)).
It is impossible to use the exact viscosity expression (Eq. (5)) because there is a singular-
ity at zero shear rate. Instead, Eq. (5) was used down to a threshold shear rate and the vis-
cosity was considered constant (at the value of the threshold viscosity) for lower shear rates.
This threshold was moved closer and closer to zero shear in successive simulations to confirm
that the velocity profile was converging to the analytical solution. Figure 5 compares the an-
alytical and simulated velocity profiles for several values of shear rate threshold. None of the
profiles deviated from the analytical solution by more than 1% at any point, and there was con-
vergence to the correct profile shape as the shear rate threshold decreased. As well as providing
evidence that the simulation can accurately capture the velocity profile of a shear-thinning fluid, this
validation suggests that the shear rate threshold does not significantly affect the shape of the velocity
profile.

VI. RESULTS AND DISCUSSION


A. Velocity and vorticity comparison
The quantitative results for the five straight tube cases are compiled in Table III, which includes
the calculated pressure gradients for Newtonian and shear-thinning fluids. Also, the normalized,
shear-thinning velocity profiles were fitted to a function of the following form:
u ∗ = 1 − (r ∗ )a . (6)
In Eq. (6), a = 2 corresponds to a parabolic profile, which was an excellent fit to all the Newtonian
cases. Larger values of a correspond to a flattened profile, so increasing values of a indicate increasing
effects of shear thinning. Finally, Table III includes the area-averaged viscosity and the Reynolds

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-9 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

TABLE III. Averaged viscosity, pressure drop, and velocity profile fits for straight tube cases. Reynolds numbers are based
on averaged viscosity, bulk velocity, and tube diameter.

dP/dx (Pa/m),
Newtonian Profile fit Averaged
Diameter (viscosity = 0.004 dP/dx (Pa/m), exponent a viscosity Reynolds
Case (cm) kg/m s) shear thinning (Eq. (6)) (kg/m s) number

Femoral 1 89.2 123.3 2.433 0.00603 123


Subclavian 0.8 120.1 162.4 2.421 0.00590 86
Splenic 0.5 762.6 770.3 2.252 0.00417 191
Hepatic 0.3 4176 4186 2.005 0.00401 223
Intercostal 0.1 11530 11530 2.007 0.00402 24

number calculated for the shear-thinning cases using this viscosity. Somewhat surprisingly in light
of previous assumptions, the deviation from Newtonian behavior increases with increasing tube
diameter. There is almost no difference between the Newtonian and shear-thinning cases for the
smallest diameters, but the pressure gradient for the shear-thinning case is 38% greater than the
Newtonian case in the largest artery. Similarly, the velocity profile distortion increases with increasing
tube diameter. There is no direct correlation with Reynolds number.
In order to evaluate the feasibility of more accurately mimicking shear-thinning flow by using
a different Newtonian viscosity, the Newtonian simulations were repeated for the three largest
arteries using the area-averaged viscosities. All three overpredicted the pressure drop calculated
in the shear-thinning simulations by 5%–10%. Previous studies did not compare shear thinning
pressure gradients to Newtonian pressure gradients with area averaged viscosity. However, the trend
of increased pressure gradient for blood compared to Newtonian fluid with blood’s asymptotic
viscosity is supported by the numerical data of Fang and Owens28 and the experiment of Thurston.59
The velocity fields for the curved tube simulations are more complex. Contour plots of some
representative cross sections of the axial and secondary velocities are shown in Figures 6 and 7,
respectively.
The flow is symmetric in the direction perpendicular to both the elbow’s radial vector and the
streamwise direction, so the top half of each plotted cross section is a Newtonian case and the bottom
half is the corresponding shear-thinning case.
All velocities are normalized by the bulk velocity, which is 0.7 m/s for the high Reynolds number
case and 0.07 m/s for the low Reynolds number case. Figure 7 shows contour plots of secondary
flow velocity magnitude instead of vector plots because the contours make it easier to compare
Newtonian and shear thinning cases visually.
Contours of axial vorticity magnitude are plotted in the same format in Figure 8.
In order to allow easier comparison between the Newtonian and shear thinning cases, the
sign of the vorticity in the bottom half of the tube cross section is reversed because the axial
vorticity distributions are antisymmetric across the white horizontal lines in Figure 8. The vorticity
is normalized by the tube diameter and the bulk inlet velocity.
Figure 6 shows that in all cases, the gradients of axial velocity are lower and the velocity profiles
are more blunted for the shear thinning flows. Peak velocities are lower for the shear thinning case
in every instance except the end of the low Reynolds number 90◦ elbow, where the Newtonian and
shear thinning axial velocity profiles have significantly different shapes. In every case, the location of
the peak velocity is closer to the outside of the elbow in the Newtonian case than the shear-thinning
case. The differences in shape are more pronounced in the 90◦ cases, particularly farther into the
elbow, where stronger secondary flows have developed.
Figure 7 shows that for both viscosity distributions, peak secondary flow magnitudes are a much
higher percentage of the axial bulk velocity in the high Reynolds number cases. Normalized by the
axial bulk velocity, the secondary flow strength doubles from low to high Reynolds number in the
90◦ cases and increases by about 50% in the 30◦ cases. The secondary flows are uniformly weaker in
the shear-thinning simulations, and these differences are more pronounced at high Reynolds number.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-10 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

FIG. 6. Axial velocity contours for curved tube simulations. For each case, the top half is the Newtonian simulation and the
bottom is the shear-thinning simulation.

Particularly notable is the region of high secondary flow near the center of the tube which is present
in the high Reynolds number, Newtonian, 90◦ elbow and absent in the corresponding shear thinning
data. Just as with the axial velocities, the peak secondary flow speeds are generally shifted farther
towards the outside of the elbow for the Newtonian flows.
It is clear from Figure 8 that the shear-thinning data are uniformly lower in axial vorticity
magnitude than their Newtonian counterparts. In the low Reynolds number cases, the vorticity
contours are basically the same shape between the Newtonian and shear thinning simulations, but
this is not always true at higher Reynolds number. The high Reynolds number data in the 90◦ elbow
show particularly drastic differences between the two flows; the shear-thinning simulation is missing
in a strong vortex that appears near the inner wall of the elbow in the 60◦ plane and migrates towards
the center of the tube downstream.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-11 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

FIG. 7. In-plane velocity magnitude contours for curved tube simulations. For each case, the top half is the Newtonian
simulation and the bottom is the shear-thinning simulation.

One useful metric for quantitatively comparing the Newtonian and shear thinning simulations
is the momentum distortion parameter, defined in Eq. (7):

2
u ax dA
A
D= − 1. (7)
u 2bulk A
Here, A is the cross-sectional area, uax is the pointwise axial velocity, and ubulk is the bulk inlet
velocity. The momentum distortion parameter (henceforth called D) is a measure of how much the
velocity profile differs from uniform flow. It is zero for a uniform flow and 1/3 for a parabolic profile.
Figures 9 and 10 show D plotted as a function of angle for the 30◦ and 90◦ elbows, respectively.
In these and the following plots, quantities are not plotted in the last 10% of the elbow because the
dominant flow feature in this region is adjustment to the straight tube section downstream of the
elbow. It is difficult to separate the effect of this adjustment from the effect of shear-thinning.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-12 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

FIG. 8. Axial vorticity contours for curved tube simulations. For each case, the top half is the Newtonian simulation and the
bottom is the shear-thinning simulation. The negative of the shear-thinning solution is plotted to allow for easier comparison.

FIG. 9. Momentum distortion parameter vs. turning angle in a 30◦ elbow.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-13 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

FIG. 10. Momentum distortion parameter vs. turning angle in a 90◦ elbow.

In both the 30◦ and 90◦ elbows, D decreases more over the elbow for the high Reynolds number
cases than the low Reynolds number ones. This is probably due to the relatively stronger secondary
flows in the high Reynolds number cases, which make the velocity profile more uniform by increasing
radial mixing.
In the 30◦ elbow, the distortion parameter is nearly constant for the low Reynolds number cases
but decreases monotonically in the high Reynolds number cases. The 90◦ elbow distortion parameters
are more complex, and some key features differ between the Newtonian and shear-thinning cases.
For example, D for the Newtonian low Reynolds number case decreases monotonically while the
corresponding shear thinning D decreases and then increases. Also, the high Reynolds number shear-
thinning distortion parameter decreases monotonically while the Newtonian curve has an inflection
point and local maximum around 60◦ from the inlet. The overall conclusion that can be reached is
that the axial velocity distribution is sensitive to the strength and orientation of the secondary flows,
which are themselves highly sensitive to the fluid rheology and the Reynolds number.
There is no analog to the distortion parameter for secondary flows and axial vorticity, so to
compare these variables quantitatively, the following difference parameter was defined:
1  |Z N ewt − Z ST |
Z = . (8)
N A Z nor m

Here, Z is a flow parameter, A is the cross sectional area of the tube, and N is the number of
points surveyed in that area. Conceptually, the difference parameter is the percentage difference in
parameter Z between the Newtonian and shear thinning simulations, averaged over the cross section
of the tube. This difference metric was chosen because it is constrained to be positive and does not
emphasize outliers as much as RMS difference does. The difference parameters for the in-plane
velocity magnitude and axial vorticity are plotted in Figures 11 and 12. For the velocity plot, Znorm
was the inlet bulk velocity, and for the vorticity plot, it was the inlet bulk velocity divided by the
tube diameter. In these plots, Z is plotted against streamwise distance along the tube instead of
angle in order to compare the 30◦ and 90◦ cases more easily.
Figures 11 and 12 provide an estimate of how much pointwise error can be expected in the
secondary flows and vorticity of a Newtonian simulation meant to approximate a shear-thinning
flow. Cross-referencing Figures 11 and 12 with Figure 7 shows both the secondary velocity and axial
vorticity difference parameters are highly correlated with the normalized strength of the secondary
flows. The difference parameters for the 30◦ cases hit a maximum about 75% of the way through
the elbow and then decrease downstream with the strength of the secondary flows.
The difference parameters for the 90◦ cases increase monotonically, which may reflect the
presence of relatively strong secondary flows at the outlet of these cases (shown in Figure 7). All high
Reynolds number cases have significantly greater difference parameters than their corresponding

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-14 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

FIG. 11. Secondary velocity magnitude difference parameter vs. arclength for curved tube simulations.

low Reynolds number cases, a trend which may also reflect increased secondary flow strength in
these cases.
However, it may be more important to examine the magnitude of the difference parameters
relative to the maximum strength of the secondary flow and vorticity, since this comparison gives an
estimate of the percentage difference between features of flows with different viscosity distributions.
In Figure 11, the low and high Reynolds number cases exhibit similar difference parameters com-
pared to their maximum secondary velocity magnitudes, peaking at 15% (30◦ elbow, low Re), 19%
(90◦ elbow, low Re), 14% (30◦ elbow, high Re), and 15% (90◦ elbow, high Re). However, Figure 12
shows that the low Reynolds number vorticity difference parameters are a significantly higher per-
centage of the flow’s maximum vorticity. They peak at 25% (30◦ elbow) and 24% (90◦ elbow) of their
corresponding maximum vorticity magnitudes compared to 15% (30◦ elbow) and 16% (90◦ elbow)
for the high Reynolds number cases. It is important to note these values are a lower bound on the
percentage difference between the two flows, since the difference parameter is averaged over the
cross section and the peak secondary velocity or vorticity occurs at only one point in the domain.

B. Curved tube pressure comparison


The last important variable that must be addressed for the curved tube cases is pressure.
Comparing pressure data from the curved tube cases is somewhat complicated because in a curved

FIG. 12. Axial vorticity difference parameter vs. arclength for curved tube simulations.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-15 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

FIG. 13. Average pressure vs. turning angle in a 90◦ elbow.

tube flow, the pressure field is 3D and nonuniform over the tube cross section. One interesting
integral measure is the pressure averaged over a cross section perpendicular to the axial flow. This
quantity is plotted for all curved tube cases in Figures 13 and 14. For all cases, the flow was set
to exit the tube at atmospheric pressure, so pressures are measured relative to the outlet. The data
are normalized by the dynamic pressure at the inlet, 1/2 ρ × ubulk 2 . The plots show that for all
cases, pressure gradient magnitude increases over the length of the elbow, but this feature is more
pronounced for the 90◦ elbow cases. This is to be expected because the geometry of the 30◦ elbow
is more similar to that of a straight pipe, which would have a constant pressure gradient.
The most striking feature of Figures 13 and 14 is that the shear-thinning flow has a consistently
higher pressure gradient than the corresponding Newtonian case for both the 30◦ and 90◦ elbows at
low Reynolds number, but this trend is reversed at high Reynolds number. A possible explanation
for this is a trade-off between the viscosity and secondary flow differences between the Newtonian
and shear-thinning cases. At low Reynolds number, shear rates are relatively low, so on average,
the shear-thinning flow has a high viscosity relative to the Newtonian flow, which leads to higher
pressure gradients.
Also, it is clear from Figures 13 and 14 that the pressure drop along the length of the tube is much
larger relative to the dynamic pressure for the low Reynolds number cases than the high Reynolds
number ones. This means differences in secondary flows between the Newtonian and shear-thinning

FIG. 14. Average pressure vs. turning angle in a 30◦ elbow.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-16 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

0.06
Brooks et al 1970
Chien and Usami 1966

dynamic viscosity (kg/m*sec)


0.05
Chien 1970
Zydney et al 1991
0.04 Gudmundsson et al 1994
Brooks et al 1970 fit
0.03 Chien 1970 fit

0.02

0.01

0 -1 0 1 2 3
10 10 10 10 10
shear rate (sec -1)

FIG. 15. Shear rate vs. blood viscosity data35, 60–63 and two fits.

cases would make a relatively smaller difference to the static pressure curves for the low Reynolds
number cases. At high Reynolds number, the reverse is true: shear rates are high and the total pressure
drop is significantly less than the inlet dynamic pressure. This means the average viscosities of the
Newtonian and shear-thinning flows are much closer and viscosity difference likely does not play a
major ole in distinguishing the pressure distributions of the two flows. However, the Newtonian cases
have stronger secondary flows that develop along the elbow, so a higher pressure gradient develops
because more static pressure must be converted to dynamic pressure. This hypothesis is supported
by the fact that the high Reynolds number pressure curves for Newtonian and shear thinning flow
are closer in the 30◦ cases than the 90◦ cases because the 90◦ cases have stronger secondary flows.

C. Blood model generality


In order to examine the sensitivity of the results to the specific viscosity data chosen, the high
and low Reynolds number 90◦ cases were re-run using a blood viscosity fit to the data of Chien.60
This fit is given by Eq. (9) and is shown in Figure 15:

−0.0236s + 0.4353
μ= for 0.17 s−1 < s < 700 s−1 ,
s 2 + 6.926s + 17.76
(9)
μ = 0.055 for s < 0.17 s−1 ,
μ = 0.004 for s > 700 s−1 .

Figure 16 compares axial velocity contours of the two blood models. These plots are of the same
format as Figures 7–9, with the Chien fit on the top half of each cross section and the Brooks et al.35
fit on the bottom half. The plots show that the two models produce very similar results, but the model
fit to the data of Chien produces slightly flattened velocity profiles compared with the model fit to
Brooks et al.35 This difference is more pronounced at high Reynolds number, but the two models
never disagree by more than 5%, and their qualitative agreement is good everywhere.
For the sake of brevity, comparison of secondary flow and vorticity contours for the two
blood models are not shown here. The models agreed qualitatively well in these variables and their
percentage differences were always equal to or smaller than 5%. The model fit to the data of Chien60
produced slightly weaker secondary flows and slightly less vorticity than the model fit to the data of
Brooks et al.35 The similarity between results from the two blood viscosity models suggests that the
comparison between Newtonian and shear-thinning flow in this study is qualitatively independent
of the experimental blood viscosity dataset and quantitatively could vary by no more than a few
percent.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-17 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

FIG. 16. Axial velocity contours for curved tube simulations. In each plot, the top half is from a model fit to the data of
Chien et al. and the bottom half is from a model fit to the data of Brooks et al.

VII. CONCLUSIONS
It often is assumed that non-Newtonian effects on blood flow are minimal in large blood vessels
and substantial in the smallest blood vessels, where the blood cell sizes are comparable to the vessel
dimensions. The results of this study show that the actual situation is not nearly so straightforward.
Simulations were performed for simple flow geometries to compare Newtonian flow to flow with a
shear thinning fluid. The latter uses an empirical relationship between blood viscosity and shear rate
that has been shown in previous experiments to be representative of blood rheology.
Simulations were performed using flow parameters characteristic of five different major arteries
in the human body. The test geometries, a straight round tube and two different curved round
tubes, are far simpler than actual blood vessels, but they capture some important effects: first of all,
the notion that shear thinning effects monotonically decrease with increasing tube diameter is not
supported by these simulations. This is apparent from the straight tube simulations, where even the
large artery cases had significant differences between Newtonian and shear thinning simulations.
Second, shear thinning flows in straight tubes have flattened velocity profiles and increased pressure
drops relative to Newtonian flows with high shear rate blood viscosity. Finally, in curved tubes,
shear thinning flows have weaker secondary flows and less axial vorticity than their Newtonian
counterparts. Cross-sectionally averaged secondary flow differences between shear thinning and
Newtonian simulations range from 15% to 19% of the maximum secondary flows, and average
vorticity differences range from 15% to 25% of the maximum vorticity. Pointwise errors in both
variables are often greater.
These conclusions are important because real arterial flows have substantial secondary flows
which can be generated by curvature of the tube, varying cross sections, or surface protrusions.
Arterial systems also contain bifurcations, which cause secondary flows. The curved tube geometries
in the present study were studied specifically to assess the effects of shear thinning on a flow in
which there are substantial secondary flows. These results give a general idea of the level of effects
that may be expected in real arteries. However, it should be clear that the observed changes are
specific for the geometries and parameters studied.
This work could be expanded by simulating pulsatile flow in curved tubes. Many studies have
examined fully developed pulsatile flow in straight tubes, but unsteady, developing flow in curved
tubes is of greater physiological relevance, and blood vessel elasticity may also play an important
role. Comparing pulsatile flow simulations with the current work would provide a more complete

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-18 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

picture of the differences between Newtonian and shear thinning tube flows. In order to do this, the
current viscosity model would have to be modified to include viscoelastic behavior, which is more
significant in unsteady flows.
1 J. P. Ku, M. T. Draney, F. R. Arko, W. A. Lee, F. P. Chan, N. J. Pelc, C. K. Zarins, and C. A. Taylor, “In vivo validation of
numerical prediction of blood flow in arterial bypass grafts,” Ann. Biomed. Eng. 30, 743 (2002).
2 B. N. Steele, J. Wan, J. P. Ju, T. J. Hughes, and C. A. Taylor, “In vivo validation of a one-dimensional finite-element method

for predicting blood flow in cardiovascular bypass grafts,” IEEE Trans. Bioeng. 50, 649 (2003).
3 C. A. Taylor, C. P. Cheng, L. A. Espinosa, B. T. Tang, D. Parker, and R. J. Herkens, “In vivo quantification of blood flow

and wall shear stress in the human abdominal aorta during lower limb exercise,” Ann. Biomed. Eng. 30, 402 (2002).
4 M. S. Olufsen, C. S. Peskin, W. Y. Kim, E. M. Pedersen, A. Nadim, and J. Larsen, “Numerical simulation and experimental

validation of blood flow in arteries with structured-tree outflow conditions,” Ann. Biomed. Eng. 28, 1281 (2000).
5 N. Shahcheraghi, H. A. Dwyer, A. Y. Cheer, A. I. Barakat, and T. Rutaganira, “Unsteady and three-dimensional simulation

of blood flow in the human aortic arch,” J. Biomech. Eng. 124, 378 (2002).
6 T. W. Taylor and T. Yamaguchi, “Three-dimensional simulations of blood flow in an abdominal aortic aneurysm—Steady

and unsteady flow cases,” J. Biomech. Eng. 116, 89 (1994).


7 W. W. Nichols and M. F. O’Rourke, Blood Flow in Arteries Theoretical, Experimental, and Clinical Principles (Oxford

University Press, New York, 1998).


8 D. Schneditz, F. Rainer, and T. Kenner, “Viscoelastic properties of whole-blood—Influence of fast sedimenting red-blood-

cell aggregates,” Biorheology 24, 1 (1987).


9 M. Kaibara and E. Fukada, “The effect of steady flow on transient viscoelastic behavior of blood,” Biorheology 18, 3–6

(1981).
10 E. Fukada and M. Kaibara, “Viscoelastic study of aggregation of red-blood-cells,” Biorheology 17, 1–2 (1980).
11 F. Stoltz, S. Gaillard, and M. Lucius, “A study of the viscoelastic properties of blood in transient flow,” J. Biomech. 13,

341–346 (1980).
12 J. L. Gennisson, S. Lerouge, and G. Cloutier, “Assessment by transient electrography of the viscoelastic properties of

blood during clotting,” Ultrasound Med. Biol. 32, 1529 (2006).


13 J. P. Mills, L. Qie, M. Dao, C. T. Lim, and S. Suresh, “Nonlinear elastic and viscoelastic deformation of the human red

blood cell with optical tweezers,” Mech. Chem. Biosyst. 1, 3 (2004).


14 A. W. L. Jay and P. B. Canham, “Viscoelastic properties of human red blood cell membrane. II. Area and volume of

individual red cells entering a micropipette,” Biophys. J. 17, 169 (1977).


15 A. W. L. Jay, “Viscoelastic properties of human red blood cell membrane. I. Deformation, volume loss, and rupture of red

cells in micropipettes,” Biophys. J. 13, 1166 (1973).


16 G. B. Thurston, “Viscoelastic properties of blood and blood analogs,” in Advances in Hemodynamics and Hemorheology,

edited by T. V. How (JAI Press Inc, Greenwich, CT, 1996).


17 H. A. G. Rojas, “Numerical implementation of viscoelastic blood flow in a simplified arterial geometry,” Med. Eng. Phys.

29, 491–496 (2007).


18 M. Joly, C. Lacombe, and J. C. Lelievre, “Tentative application of the tangent simple system method to the study of the

viscoelastic behavior of blood,” Biorheology 20, 5 (1983).


19 D. Lerche, G. Vlastos, B. Koch, M. Pohl, and K. Affeld, “Viscoelastic behavior of human blood and polyacrylamide model

fluids for heart valve testing,” J. Phys. III 3, 1283–1289 (1993).


20 T. Bodnar, K. R. Rajagopal, and A. Sequeira, “Simulation of the three-dimensional flow of blood using a shear-thinning

viscoelastic fluid model,” Math. Models Nat. Phenom. 6, 5 (2011).


21 A. R. Pries, D. Neuhaus, and P. Gaehtgens, “Blood viscosity in tube flow: Dependence on diameter and hematocrit,” Am.

J. Physiol. Heart Circ. Physiol. 263, H1770 (1992).


22 C. Alonso, A. R. Pries, and P. Gaehtgens, “Time-dependent rheological behavior of blood at low shear in narrow vertical

tubes,” Am. J. Physiol. Heart Circ. Physiol. 265, H553 (1993).


23 A. E. Medvedev and V. M. Fomin, “Two-phase blood flow model in large and small vessels,” Dokl. Phys. 56, 610 (2011).
24 M. Sharan and A. S. Popel, “A two-phase model for flow of blood in narrow tubes with increased effective viscosity near

the wall,” Biorheology 38, 415 (2001).


25 P. K. Nair, J. D. Hellums, and J. S. Olson, “Prediction of oxygen transport rates in blood flowing in large capillaries,”

Microvasc. Res. 38, 269 (1989).


26 V. Seshadri and M. Y. Jaffrin, “Anomalous effects in blood flow through narrow tubes: A model,” INSERM-Euromech.

71, 265 (1977).


27 J. K. W. Chesnutt and J. S. Marshall, “Structural analysis of red blood cell aggregates under shear flow,” Ann. Biomed.

Eng. 38, 714 (2010).


28 J. Fang and R. G. Owens, “Numerical simulations of pulsatile blood flow using a new constitutive model,” Biorheology

43, 637 (2006).


29 P. M. Vlahovska, T. Podgorski, and C. Misbah, “Vesicles and red blood cells in flow: From individual dynamics to

rheology,” C. R. Phys. 10, 775 (2009).


30 E. R. Damiano, “Blood flow in microvessels lined with a poroelastic wall layer,” in Poromechanics, edited by J. F. Thimus,

Y. Abousleiman, A. H. D. Cheng, O. Coussy, and E. Detournay (Balkema, Leiden, 1998), pp. 403–408.
31 K. Sriram, A. G. Tsai, P. Cabrales, F. Meng, S. A. Acharya, D. M. Tartakovsky, and M. Intaglietta, “PEG-albumin

supraplasma expansion is due to increased vessel wall shear stress induced by blood viscosity shear thinning,” Am. J.
Physiol. Heart Circ. Physiol. 302, H2489 (2012).
32 F. H. El-Khatib and E. R. Damiano, “Linear and nonlinear analysis of pulsatile blood flow in a cylindrical tube,” Biorheology

40, 503 (2003).

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
073104-19 E. M. Cherry and J. K. Eaton Phys. Fluids 25, 073104 (2013)

33 S. McGinty, S. McKee, and R. McDermott, “Analytic solutions of Newtonian and non-Newtonian pipe flows subject to a
general time-dependent pressure gradient,” J. Non-Newtonian Fluid Mech. 162, 54 (2009).
34 M. A. Moyers-Gonzalez, R. G. Owens, and J. Fang, “On the high frequency oscillatory tube flow of healthy human blood,”

J. Non-Newtonian Fluid Mech. 163, 45 (2009).


35 D. E. Brooks, J. W. Goodwin, and G. V. F. Seaman, “Interactions among erythrocytes under shear,” J. Appl. Physiol. 28,

172 (1970).
36 R. E. N. Shehada, R. S. C. Cobbold, and P. A. J. Bascom, “Ultrasound methods for investigating the non-Newtonian

characteristics of whole blood,” IEEE Trans. Ultrason., Ferroelectr., Freq. Control 41, 96 (1994).
37 T. Osada and G. Radegran, “Alterations in the rheological flow profile in conduit femoral artery during rhythmic thigh

muscle contractions in humans,” Jpn. J. Physiol. 55, 19 (2005).


38 T. Kiserud, L. R. Hellevik, and M. A. Hanson, “Blood velocity profile in the ductus venosus inlet expressed by the

mean/maximum velocity ratio,” Ultrasound Med. Biol. 24, 1301 (1998).


39 C. J. Ade, R. M. Broxterman, B. J. Wong, and T. J. Barstow, “Anterograde and retrograde blood velocity profiles in the

intact human cardiovascular system,” Exp. Physiol. 97, 849 (2012).


40 Y. Amano, R. Takagi, Y. Suzuki, T. Sekine, S. Kumita, and M. van Cauteren, “Three-dimensional velocity mapping of

thoracic aorta and supra-aortic arteries in Takayasu arteritis,” J. Magn. Reson Imaging 31, 1481 (2010).
41 C. Marquis, J. J. Meister, V. Mirkovitch, C. Depeursinge, E. Mooser, and R. Mosimann, “Femoral blood flow determination

with a multichannel digital pulsed Doppler: An experimental study on anesthetized dogs,” Vasc. Endovasc Surg. 17, 95
(1983).
42 S. Ricci, G. Urban, P. Vergani, M. J. Paidas, and P. Tortoli, “Blood flow imaging in maternal and fetal arteries and veins,”

Accoust. Imaging 29, 87 (2009).


43 T. W. A. Huisman, P. A. Stewart, and J. W. Wladimiroff, “Ductus venosus blood flow velocity waveforms in the human

fetus—A Doppler study,” Ultrasound Med. Biol. 18, 33 (1992).


44 M. B. Histand, C. W. Miller, and F. R. Mcleod, Jr., “Transcutaneous measurement of blood velocity profiles and flow,”

Cardiovasc. Res. 7, 703 (1973).


45 J. J. Bishop, P. R. Nance, A. S. Popel, M. Intaglietta, and P. C. Johnson, “Effect of erythrocyte aggregation on velocity

profiles in venules,” Am. J. Physiol. Heart Circ. Physiol. 280, H222 (2001).
46 G. R. Cokelet and H. L. Goldsmith, “Decreased hydrodynamic resistance in the two-phase flow of blood through small

vertical tubes at low flow rates,” Circ. Res. 68, 1 (1991).


47 G. J. Tangelder, D. W. Slaaf, A. M. Muijtjens, T. Arts, M. G. oude Egbrink, and R. S. Reneman, “Velocity profiles of blood

platelets and red blood cells flowing in arterioles of the rabbit mesentery,” Circ. Res. 59, 505 (1986).
48 C. Alonso, A. R. Pries, O. Kiesslich, D. Lerche, and P. Gaehtgens, “Transient rheological behavior of blood in low-shear

tube flow: Velocity profiles and effective viscosity,” Am. J. Physiol. Heart Circ. Physiol. 268, H25 (1995).
49 S. Agrawal, V. K. Srivastava, G. Jayaraman, and K. D. P. Nigam, “Power law fluids in a circular curved tube. Part III.

numerical simulation of laminar flow,” Polym.-Plast. Technol. Eng. 33, 357 (1994).
50 F. J. H. Gijsen, E. Allanic, F. N. van de Vosse, and J. D. Janssen, “The influence of the non-Newtonian properties of blood

on the flow in large arteries: Unsteady flow in a 90◦ curved tube,” J. Biomech. 32, 705 (1999).
51 M. Huonker, A. Schmid, A. Schmidt-Trucksass, D. Grathwohl, and J. Keul, “Size and blood flow of central and peripheral

arteries in highly trained able-bodied and disabled athletes,” J. Appl. Physiol. 95, 685 (2003).
52 L. A. Silveira, F. B. Silveira, and V. P. Fazan, “Arterial diameter of the celiac trunk and its branches. Anatomical study,”

Acta Cir. Bras. 24, 43 (2009).


53 S. Sato, K. Ohishi, S. Sugita, and K. Okuda, “Splenic artery and superior mesenteric artery blood flow: Nonsurgical Doppler

US measurement in healthy subjects and patients with chronic liver disease,” Radiology 164, 347 (1987).
54 Y. Kito, M. Nagino, and Y. Nimura, “Doppler sonography of hepatic arterial blood flow velocity after percutaneous

transhepatic portal vein embolization,” Am. J. Roentgenol. 176, 909 (2001).


55 R. P. Da Rocha, A. Vengjer, A. Blanco, P. T. Carvalho, M. L. Mongon, and G. J. Fernandes, “Size of the collateral intercostal

artery in adults: Anatomical considerations in relation to thoracocentesis and thoracoscopy,” Surg. Radiol. Anat. 24, 23
(2002).
56 T. Koyanagi, N. Kawaharada, Y. Kurimoto, T. Ito, T. Baba, M. Nakamura, A. Watanebe, and T. Higami, “Examination of

intercostal arteries with transthoracic Doppler sonography,” Echocardiogr. 27, 17 (2010).


57 P. H. M. Bovendeerd, A. A. van Steenhoven, F. N. van de Vosse, and G. Vossers, “Steady entry flow in a curved pipe,”

J. Fluid Mech. 177, 233 (1987).


58 T. J. Barth and D. Jespersen, “The design and application of upwind schemes on unstructured meshes,” AIAA Paper No.

89-0366, 1989.
59 G. B. Thurston, “Elastic effects in pulsatile blood flow,” Microvasc. Res. 9, 145 (1975).
60 S. Chien, “Shear dependence of effective cell volume as a determinant of blood viscosity,” Science 168, 977 (1970).
61 A. L. Zydney, J. D. Oliver III, and C. K. Colton, “A constitutive equation for the viscosity of stored red cell suspensions:

Effect of hematocrit, shear rate, and suspending phase,” J. Rheol. 35, 1639 (1991).
62 S. Chien, S. Usami, H. M. Taylor, J. L. Lundberg, and M. I. Gregersen, “Effects of hematocrit and plasma proteins on

human blood rheology at low shear rates,” J. Appl. Physiol. 21, 81 (1966).
63 A. Gudmundsson, A. Oden, and A. Bjelle, “On whole blood viscosity measurements in healthy individuals and in

rheumatoid arthritis patients,” Biorheology 31, 407 (1994).

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
171.67.216.21 On: Fri, 03 Apr 2015 18:52:18
View publication stats

You might also like