Rheological and yield stress measurements of non-Newtonian uids using a

Marsh Funnel
Matthew T. Balhoff , Larry W. Lake, Paul M. Bommer, Rebecca E. Lewis, Mark J. Weber, Jennifer M. Calderin
Petroleum and Geosystems Engineering, University of Texas at Austin, 1 University Station C0300 Austin, TX 787120228, USA
a b s t r a c t a r t i c l e i n f o
Article history:
Received 30 September 2010
Accepted 10 April 2011
Available online 29 April 2011
Keywords:
Marsh Funnel
Yield stress
Rheology
Drilling mud
Polymer
Non-Newtonian
Accurate and simple techniques for measurement of uid rheological properties are important for eld
operations in the oil industry, but existing methods are relatively expensive and the results can be subjective.
This is particularly true for measurements of uid yield stress, which are notoriously difcult to obtain. Marsh
Funnels are popular quality-control tools used in the eld for drilling uids and they offer a simple, practical
alternative to viscosity measurement. In the normal measurements, a single point (drainage time) is used to
determine an average viscosity; little additional information is extracted regarding the non-Newtonian
behavior of the uid.
Here, a new model is developed and used to determine the rheological properties of drilling muds and other
non-Newtonian uids using data of uid volume collected from a Marsh Funnel as a function of time. The
funnel results for viscosity and shear-thinning index compare favorably to the values obtained from a
commonly-used Fann 35 viscometer. More importantly, an objective, static method for determining yield
stress is introduced, which has several advantages over dynamic, extrapolation techniques used for
rheometer data.
Published by Elsevier B.V.
1. Introduction
1.1. Non-Newtonian uids in the petroleum industry
Hydrocarbon production uses many uids that are rheologically
complex. Among these are cement, drilling muds, aqueous solutions
of water-soluble polymer, and, of course, crude oil itself.
Drilling uids can be air or water, but most commonly they are
muds or suspensions of solids in an aqueous or oleic uid. The solids
are suspended with one or more surfactants. The solids are to provide
weight to the mud for pressure control, the main function of muds,
but muds also lubricate the drill, carry drilling cuttings to the surface
and cool the bit. Most muds are water-based as is the type used in this
study. When fresh water is the liquid base, bentonite is the clay used
for its superior properties necessary to achieve the goals stated for
drilling mud. Salt water mud can be created using bentonite that has
been pre-hydrated with fresh water so long as the salinity is not much
more than that of sea water.
Drilling mud exhibits several important rheological properties
(Bourgoyne et al., 1991). The viscosity or consistency index of a mud is
a measure of ow resistance. Therefore viscosity should be as small as
possible to limit friction pressure. However a certain amount of
viscosity is required to improve the solids carrying capacity of the
mud. If viscosity is too small, the mud may be unable to suspend
drilled solids at the desired pump rate. This requires the pumps to be
run faster to continue to circulate drilled solids out of the well. If
viscosity is too high, an excessive pump pressure will be required to
circulate the mud at the desired rate. Higher than necessary pump
pressure is an added strain on the pumps and piping and an added
pressure in the bore hole that can lead to well bore stability problems.
Water-soluble polymers are also used in drilling uids to improve
the ability of muds to lift cuttings, but they are also used as fracturing
uids to improve the removal of solids after fracturing, and in
enhanced oil recovery. Two common polymers used are xanthan gum
(hereafter just xanthan) and partially hydrolyzed polyacrylamide
(HPAM). These are also used in the current study.
These non-Newtonian uids (drilling muds and polymers) may
also exhibit a yield stress (or gel strength). For drilling operations, the
higher the yield stress the more pump pressure will be required to
initiate circulation. The yield stress can also be a desirable property
because it will suspend the drilled solids and prevent or slow them
from slipping back to the bottom of the hole during periods when
there is no circulation. Fluid yield stress in fracturing uids can help
carry and suspend proppant, but can also make cleanup difcult (May
et al., 1997; Balhoff and Miller, 2005).
Journal of Petroleum Science and Engineering 77 (2011) 393402
Corresponding author. Tel.: +1 512 471 3246; fax: +1 512 471 9605.
E-mail address: Balhoff@mail.utexas.edu (M.T. Balhoff).
0920-4105/$ see front matter. Published by Elsevier B.V.
doi:10.1016/j.petrol.2011.04.008
Contents lists available at ScienceDirect
Journal of Petroleum Science and Engineering
j our nal homepage: www. el sevi er. com/ l ocat e/ pet r ol
1.2. Rheological models
Purely viscous, non-Newtonian uids are often classied using
constitutive models relating the shear stress to the shear rate
(Table 1). Many of these models (e.g. power-law, Carreau) account
for the shear-thinning (decrease in viscosity with shear rate) behavior
observed in many non-Newtonian uids. For shear-thinning uids,
the power-law index, n, is less than one. The uid is Newtonian and
the model reduces to Newton's law if n equals one.
Some uids exhibit a yield stress, requiring a minimum stress to
initiate ow. Belowthe yield stress the material is solid-like and has an
innite viscosity. The solid-like behavior is typically a result of a three-
dimensional microstructure at low stresses (Carreau et al., 1997).
Above the yield stress the material deforms as a uid and the viscosity
is a function of shear rate. Many pastes, foodstuffs, gels, and drilling
muds have a yield stress. The simplest yield-stress model is the
Bingham model, in which the relationship between shear stress and
shear rate is linear, with the yield stress dened as the extrapolated y-
axis intercept. A more general model is the HerschelBulkley
constitutive equation, which has the shear-thinning (or shear-
thickening) behavior of power-law uids and the yield-stress effect
of the Bingham model (see Table 1). The HerschelBulkley model
reduces to the ideal Bingham model for the special case where n=1,
and to the power-law model for no yield stress (
0
=0). Although
more complex models may describe the rheology of specic uids
better, the Bingham and HerschelBulkley models are widely used
because of their mathematical simplicity.
To mathematically describe the rheology of a uid, a constitutive
equation must be chosen and the empirical constants (e.g. m, n,
0
)
must be determined experimentally. Typically, experiments are
performed using a rheometer with a couette, parallel plate, or cone
and plate geometry. Shear stress (or viscosity) is measured dynam-
ically as a function of shear rate and a best-t match to the data
determines the model constants. Accurate results are often obtained
for the shear-thinning (n) and consistency (m) index provided
enough data points are extracted over a wide range of shear rates.
Although highly accurate rheometers are available, simple lab
rheometers are commonly used in the oileld, perhaps the most
popular of which is the Fann 35 viscometer (Fann, 2010). The Fann 35
is small, light, relatively inexpensive, and easy to use. It is often used
to obtain quick rheological measurements of drilling muds and other
oileld uids. The couette geometry allows for specication of
revolutions per minute (RPM), which is proportional to shear rate,
and measures the resulting torque, which is proportional to shear
stress. Although a constitutive model can be t to the data, only six
shear rates (in a limited range) are obtainable, making parameter
estimation (especially yield stress) difcult. Screen factor devices
(Lake, 1989) are also popular for measuring the viscosity of polymers
relative to brine, but only one data point is obtained. Curves of
viscosity as a function of shear rate are not possible with the screen
factor device.
1.3. Yield stress measurement
The uid yield stress is historically difcult to measure (Nguyen
and Boger, 1983; Zhu et al., 2001; Balhoff, 2005) in any rheometer.
Measurement of yield stress from a ow curve of shear stress versus
shear rate can give faulty results for a number of reasons. First, the
yield stress must be determined by extrapolation of the curve to a
shear rate of zero, since the lowest shear measurement is nite and
limited by the rheometer. Second, measurements performed at low
shear rates may be inaccurate because of slip at the walls (Nguyen and
Boger, 1983). Finally, the yield stress measured may be a dynamic
yield stress and not the stress required to break the three-dimensional
structure of the material and initiate ow(Carreau et al., 1997). Other
tests such as creep, oscillatory shear, and stress ramp can be used to
measure yield stress, but they can also be ambiguous and inconsistent
(Nguyen and Boger, 1992; Carreau et al., 1997). Some authors even
claim that a true yield stress does not exist (Barnes and Walters, 1985)
while others claim otherwise (Peder et al., 2006; Moller et al., 2009).
Certain direct tests, such as the vane (Nguyen and Boger, 1983)
and plate (Carreau et al., 1997) method, tend to be more reliable and
less subjective than the aforementioned dynamic tests. In each case,
torque (proportional to stress) versus time (proportional to strain) is
measured. Initially, the torque increases linearly with time, demon-
strating the Hookian behavior of the material. After the linear region
the torque may continue to increase non-linearly, indicating visco-
elastic behavior. Once the torque reaches a maximum, the internal
structure breaks down and the material yields. The yield stress can be
determined from this maximum torque (Carreau et al., 1997). A
slotted plate device has been developed by Zhu et al. (2001) that is
particularly useful for suspensions that exhibit a low yield stress,
when the vane or plate may be less reliable.
1.4. Marsh Funnel
The Marsh Funnel was invented by Hallan N. Marsh in 1931
(Marsh, 1931). It is used to measure the time in seconds required to
ll a set volume of uid. (In the United States the volume is one quart.)
The owthrough the small tip at the end of the funnel is related to the
rheological properties of the uid being measured. The Marsh Funnel
viscosity is reported as seconds and used as an indicator of the
relative consistency of uids, the more viscous the uid the longer the
time to ll one quart. The calibration for Marsh Funnel time is
25 seconds per quart for fresh water. The standard Marsh Funnel is
shown in Fig. 1 (picture taken from Schlumberger). The Marsh Funnel
Table 1
Constitutive equations for rheological models and the resulting steady ow equations in a cylindrical tube of radius R and length L.
Model Constitutive equation Cylindrical tube model
Newtonian =

Q =
R
4
8 L
P
Power-law = m
n
Q =
nR
3 + 1
=n
3n + 1 2mL
1
=n
P
1
=n
Bingham
(Skelland, 1967)
=
0
+ m

Q =
R
4
8m L
P 1
4
3
2L
0
PR
_ _
+
1
3
2L
0
PR
_ _
4
_ _
HerschelBulkley
(Skelland, 1967)
=
0
+ m

n
Q =
R
3
m
1
=n

3
w
w
0

1
=n + 1 w0
2
1
=
n + 3
+
20 w0
1
=
n + 2
+

2
0
1
=
n + 1
_ _
General Fluid
(Carreau et al., 1997)
=

Q =
R
3

w
3

3
R
w
_ _
3

w
0

3
d

394 M.T. Balhoff et al. / Journal of Petroleum Science and Engineering 77 (2011) 393402
provides a simple andeffective tool to determine the relativeviscosityof
drilling mud. Here, we also use the funnel for additional oileld uids.
Although rheological properties of these uids can be measured by
conventional rheometers, a simple method is often needed. The goal of
this work is to develop such a method for determining rheological
properties of non-Newtonian uids using a Marsh Funnel. An
experiment consists of lling the funnel to a pre-specied height and
measuringthe rate at whichthe test uiddrains. Pitt (2000) developeda
model for non-Newtonianowin a MarshFunnel, but his approach was
not used to estimate the rheological parameters uniquely. The model
presented here can estimate multiple parameters for a uid uniquely.
Moreover, an objective measurement of yield stress is introduced using
the nal, non-zero uid height in the funnel. We also show that the
Marsh Funnel could be used as a tool for eld measurement of uids
used in enhanced oil recovery and fracturing as well.
2. Experimental
2.1. Fluid preparation
Many of the uids used in these experiments must be mixed before
testing. The uids that are used in this project are mineral oil, polymer
(xanthan, hydrolyzed polyacrylamide, carbopol), and bentonite. Before
adding any solid particles or polymer to the water-based uids used
here, the water was adjusted to approximately a pH of 9 by adding
droplets of 2 weight % NaOH (sodiumhydroxide). For all tests, the uid
was allowed to cool to room temperature (~19 C), the density was
measured using a density balance, and the rheology measured using a
Fann 35 viscometer. Fluid was then poured in the Marsh Funnel for the
tests. The sections below describe how each uid is prepared.
2.1.1. Mineral oil
A Newtonian uid was used as a standard test for the Marsh
Funnel experiments. The viscosities of uids using the Marsh Funnel
are often compared to the drainage time for water, but this can be
very misleading since viscosity is a very nonlinear function for
turbulent ow. Furthermore, the drainage time for low viscosity uids
suchas water are dominated by the resistance of the funnel constriction
and independent of uid viscosity (Pitt, 2000). Since the model
developed here assumes laminar ow, a higher viscosity uid than
water is needed. Mineral oil (~0.20 Pa-s) was chosen instead as the
Newtonian basis.
2.1.2. Xanthan polymer
Xanthan polymer is widely used in several industries, including
petroleum. Xanthan was chosen as a test non-Newtonian uid in
this work because it is shear-thinning, may exhibit a yield stress
(Song et al., 2006), and exhibits minimal elastic effects in
comparison to polyacrylamides (Lake, 1989). Concentrations of 1
2 wt% xanthan polymer were used in this study. The uid was mixed
vigorously for several minutes and allowed to cool (the uid heated
as a result of mixing) for several hours before tests in the rheometer
and funnel. The molecular weight of the xanthan used here is
unknown.
2.1.3. Bentonite muds
Bentonite muds are the most common used in oil drilling usually
at concentrations of 5 lb/bbl (1.4 wt%) and larger. They are well
known to be non-Newtonian and may exhibit a yield stress at high
concentrations. Here, low-gravity solids are added to simulate real-
world effects of drilling mud as it comes out of a borehole carrying
drilled cuttings. Low-gravity solids were mixed with the bentonite
mixtures for approximately 30 minutes and then allowed to cool
before using in the experimental tests. Bentonite exhibits some time-
dependent properties (including yield stress), but enough time
elapsed (hours) so that the properties were not dynamic (Gray et al.,
1980).
2.1.4. HPAM
Hydroxypolyacrylamide (HPAM) is widely used in the petroleum
industry for chemical EOR processes (Lake, 1989). The uid is shear-
thinning but also exhibits viscoelasticity, hysteresis, and other time-
dependent effects (Lake, 1989; Carreau et al., 1997). HPAM was
created by adding polymer to deionized water with 1% NaCl. The
solution was stirred for 72 hours and then ltered.
2.1.5. Carbopol
Carbopol gel is used in household and personal care products.
The polymer is well-known to be shear-thinning and is also
reported to exhibit a yield stress (Chase and Dachavijit, 2003).
Here, Carbopol 980 was obtained from Noveon Inc. The powder was
added to 3% NaCl solution and stirred for approximately 24 hours.
Ten percent NaOH was then added until the pH reached approxi-
mately 7.5 which crosslinked the gel and created a viscous, clear
solution.
2.2. Rheology tests
We used a Fann 35 viscometer to measure the uid's rheological
properties which are then compared to results obtained using the
Marsh Funnel. The Fann 35 cups are lled to the top of the dashed
line and placed on the base of the rheometer. Once the cup is
positioned, the rheometer is turned onto its highest setting of
600 rpm. The degree dial is allowed to stabilize before taking the
rst recording. Subsequent measurements at 300, 200, 100, 6, and
3 rpm are recorded as well along with the resulting dial reading. The
data are then converted to shear stress and shear rate, respectively.
Details can be found in the instruction manual (Fann, 2010). Some
uids were also tested on an ARES LS-1, controlled-rate, tempera-
ture-controlled rheometer. Plots of shear stress versus shear rate are
(15.2 cm)
6 in.
(0.475 cm)
(0.16 cm)
12 in. (30.5 cm)
2 in. (5 cm)
Fig. 1. Standard Marsh Funnel. The height of cone-portion of the funnel is 12 in.
(30.5 cm) and the diameter is 6 in. (15.2 cm). The copper tubing is 2 in. (5.08 cm) in
length and has a diameter of 3/ in. (0.48 cm).
395 M.T. Balhoff et al. / Journal of Petroleum Science and Engineering 77 (2011) 393402
generated and rheological properties (i.e. viscosity, power index,
and yield stress) are obtained through curve tting to various
rheological models (Table 1).
2.3. Marsh Funnel test
After the uid rheology is measured, the uid is placed in the
Marsh Funnel (Fig. 1). The Marsh Funnel is designed so that 1500 mL
of uid can be poured into the funnel. A small stopper is placed in the
orice at the bottomto prevent owout while the uid is poured into
the funnel. Once it reaches the bottomof the screen, this indicates that
1500 mL now rests in the funnel. The purpose of the screen is to
remove any unmixed solid particles from the rest of the uid. The
uids are allowed to rest in the funnel for a few minutes before being
tested to potentially build up gel strength.
Once the funnel is lled, a beaker is placed on a scale positioned
below the funnel. The scale is connected to a computer that records
weight versus time at intervals of 1 second. The weight is converted
to a volume using the uids density. The plug is removed from the
orice and the uid is allowed to ow until it comes to a complete
stop or completely drains. For uids that exhibit a yield stress, a
steady-state height is recorded which can be related to the yield
stress. The nal height is measured using a submersible measuring
stick. Most tests were repeated 23 times and excellent repeatability
was obtained.
Some of the more viscous uids adhere to the sides of the funnel
wall making the volume collected plus volume remaining in the static
head less than 1500 mL. For these cases, the volume versus time data
is corrected by assuming accumulation occurs uniformly along the
walls with funnel height. The corrected volume versus time curve is
then converted to a height versus time curve using the mathematical
equations developed in Section 3.
3. Model development
3.1. Model assumptions
A model for uid height (h) as a function of time (t) in the funnel is
developed and rheological properties are then determined froma best
t of the data to the model. The analysis is based on a solution to an
ordinary differential equation that is based on the following
assumptions:
1. The density of air is negligible, compared to the uid densities, as is
capillary pressure between air and the existing uid.
2. The uids being tested are incompressible.
3. There is no viscous resistance to ow in the funnel itself. This
assumption implies that pressure at the top of the nozzle region is
given by the hydrostatic head of the funnel and is veried in the
Appendix. The hydrostatic head is independent of the funnel
geometry for uids without a yield stress. The assumption also
eliminates entrance effects at the bottom of the funnel, which
could lead to inaccuracies since L/D of the capillary tube is
approximately 10.
4. Elastic effects are negligible.
5. Flowin the nozzle region is fully developed, steady-state Poiseuille
ow. The basic measurements are necessarily transient and this
assumptions means that ow is quasi-static. This assumption is
veried with calculations in the Appendix. The owis laminar in all
tests conducted here.
6. There is no loss of uids to evaporation or to the sides of the funnel
wall. While uid loss is not included directly into the ODE, it is
corrected during data reduction as described in Section 2.3 to
achieve material balance.
3.2. Model derivation
For these assumptions, a mass balance for funnel uid can be
written as:
dV
dt
= Q h 1
The volume of uid in the funnel can be written as a function of
height using the formula for a cone (V=r
2
h/3). The radius can be
related to the height using similar triangles:
r =
R
F
H
F
_ _
h 2
V =

3
R
F
H
F
_ _
2
h
3
= h
3
3
Where R
F
is the maximum funnel radius and H
F
is the maximum
height. The above equation has a coefcient =0.065 using the
dimensions of the funnel. A calibration curve for volume versus height
has shown =0.078 is more accurate and is used in this work.
Substituting Eq. (3) into the mass balance and using the chain rule
yields:

R
F
H
F
_ _
2
h
2
dh
dt
= Q h 4
Eq. (4) is an ODE that describes the height of uid in the Marsh
Funnel as a function of time. An appropriate equation for Q(h) in a
cylindrical tube is needed. The equation for ow, Q, in a capillary
(bottomof funnel) for various non-Newtonian uid models is given in
Table 1. The pressure drop is the hydrostatic head; the total uid
height being the sum of the tube length, L, and funnel height, h.
Where the shear stress at the wall of the tube is given by (see the
Appendix):

w
=
PR
2L
=
g h + L R
2L

2H
F
R
F

0
5
The goal is to determine rheological properties using the time
elapsed and height of uid displaced. For Newtonian ow, the Hagen
Poiseuille equation can be substituted into the ODE. Separation of
variables gives:

R
F
H
F
_ _
2
8L
R
4
_ _

h
h
0
h
2
g h + L
dh = t 6
Integrating Eq. (6) relates the funnel height as a function of time.
L
2
ln
h + L
h
0
+ L
_ _

1
2
h
2
2hL
_ _
h
2
0
2h
0
L
_ _ _ _
=
R
4
8 L
H
F
R
F
_ _
2
g
_ _
t 7
h=0 at t =t
f
(drainage time). Also, using the fact that:
L
2
ln
L
h
0
+ L
_ _
bbLh
0

1
2
h
2
0
8
The viscosity of a Newtonian uid can then be estimated by the
following equation and the total drainage time, t
f
, for laminar ow.

R
4
8L
H
F
R
F
_ _
2
_ _
g
Lh
0

1
_
2
h
2
0
_ _
t
f
9
For non-Newtonian uids, Q(h) is more complicated (Table 1) and
in general Eq. (4) must be integrated numerically (e.g. 4th order
396 M.T. Balhoff et al. / Journal of Petroleum Science and Engineering 77 (2011) 393402
Runge-Kutta method). The rheological parameters can be found by
tting the model solution of h(t) to the dynamic funnel data;
nonlinear regression can be used to nd the best t. Microsoft Excel's
SOLVER function was used in this work to nd the model parameters.
The funnel data is never converted to shear rate and shear stress in the
present work (although such a conversion could be performed).
Parameter estimation is performed by nding a least-squares t of the
height versus time data to the solution of the model (Eq. (4)).
The uid yield stress could be determined in the same fashion as
the other rheological properties, i.e. as a tted parameter in the least-
squares optimization. However, the yield stress is a fundamental uid
property and should be measured under static conditions (Zhu et al.,
2001) if possible. Moreover, a goal of this work is to identify a simple,
objective measurement technique for yield stress. The yield stress is
the minimum shear stress required to induce ow, or equivalently
the maximum stress that can be imposed before ow occurs. Here,
the shear stress can be related to the height of uid in the funnel; the
static height where the forces are in balance (i.e. dh/dt =0). The
derivation is given in the Appendix.

0
=
w
=
g h
ss
+ L
2L
R
+
2H
F
R
F
10
If there is no static height, the uid exhibits no yield stress or the
yield stress is smaller than can be measured with the capillary radius
of the funnel. The smallest yield stress that can be measured in the
Marsh Funnel is 12.5 Pa. Once the yield stress is determined under the
static conditions, it can be substituted directly into the ODE (Eq. (4));
the consistency index (m) and shear-thinning index (n) are then
determined from nonlinear regression by minimizing the sum of
squared errors between the model and data.
4. Results and discussion
4.1. Mineral oil (Newtonian standard)
The shear stress, shear rate data obtained fromthe rheometers can
be plotted to determine the rheological model that best describes the
uid and the best-t rheological properties for that model. The results
can be compared to those obtained using the Marsh Funnel.
Fig. 2 shows shear stress versus shear rate data for mineral oil. The
data are clearly linear validating Newtonian behavior with a viscosity
of 0.19 Pa-s (from the Fann 35 viscometer). A slightly higher viscosity
(0.20 Pa-s) was obtained using ARES rheometer. Fig. 3 is a plot of uid
height versus time in the funnel for mineral oil. An excellent match
between the data and the best-t solution to model ODE (assuming a
Newtonian uid in laminar ow) is obtained. Moreover, the viscosity,
0.18 Pa-s, is close to the values found using traditional rheometers.
Fig. 3 also shows the predicted curve based on the Fann 35 estimated
rheology, which is shifted to the right. Note that the curve in Fig. 3 is
concave down for the entire time scale of the experiment. This
indicates a high shear-thinning index (near one); other experiments
showed concave-up behavior (at least at early times) suggesting a low
shear-thinning index.
4.2. Xanthan polymer
Fig. 4a shows the rheological data for 1.1 wt% xanthan polymer. A
good t can be obtained using a power-law model and rheological
parameters n=0.10 and m=15 Pa-s
n
using the Fann 35. Some
authors (Song et al., 2006) report that xanthan exhibits a yield stress
and a HerschelBulkley model would t the data equally as well by
extrapolating the shear stress curve to zero shear rate and using
parameters n=0.51, m=0.52 Pa-s
n
, and
0
=17 Pa. The ambiguity in
the stress at zero shear rate demonstrates the difculty in yield stress
measurement using a traditional rheometer. Any number of values for

0
between 0 and 17 Pa would result in an acceptable t, each with
different HerschelBulkley parameters (n and m). Fig. 4b plots the
same rheology data as viscosity versus shear rate on a loglog plot.
The funnel data for 1.1 wt% xanthan is shown in Fig. 5. All of the
uid drains from the funnel suggesting no yield stress (or at least
below the measurable value of
0
=12.5 Pa). However, a power-law
model does not adequately t the funnel data either (signicant
systematic deviation is observed). Inspection of the shear rates in the
funnel show many values well above 1000 1/s, where a Newtonian
plateau may exist (Fig. 4b), suggesting a Carreau model (Eq. (11))
may be better suited for the data. No analytical solution is available for
owrate in a tube for Carreau uids, but the integral shown in Table 1
can be evaluated numerically if the wall shear rate (
w
) is known.
Given the shear stress at the wall (Eq. (5)), the wall shear rate can be
found from the nonlinear Eq. (12).

+
0

1 +

2
_ _
n1
2
11
Fig. 2. Rheological data for mineral oil. A Newtonian t to the data gives a viscosity of
0.19 Pa-s using the Fann 35 and 0.20 Pa-s using the ARES rheometer.
Fig. 3. Height versus time data for mineral oil. The Newtonian model gives an excellent
t to the data with a viscosity of 0.18 Pa-s. The predicted curve based on the Fann 35
estimated viscosity (0.19 Pa-s) shows a slightly longer drainage time.
397 M.T. Balhoff et al. / Journal of Petroleum Science and Engineering 77 (2011) 393402

w
=
w

12
The parameter can be converted to the consistency index, m,
used in the power-law model by:
m =
0

n1
13
The algorithm for solving the ODE for a Carreau uid requires
several additional numerical approximations; (1) provide initial guess
the Carreau rheological properties, (2) at a given stepsize in the
numerical ODE solver, calculate the wall stress using Eq. (5); (3) solve
for the wall shear rate in Eq. (12) using Newton's method; (4)
compute the integral in Table 1 numerically and calculate the
owrate, Q. Update the height for the next timestep using Q.
Fig. 5 shows a good Carreau model t to the data for 1.1 wt%
xanthan using parameters
0
=220 Pa-s,

=0.014 Pa-s, m=18 Pa-

s
n
, and n=0.12. These values now compare favorably to the values
obtained from the Fann 35 for the power-law regime (m=15 Pa-s
n
and n=0.10). One advantage of the funnel is the ability to estimate
Carreau model parameters that are relevant outside of the shear rate
range of the Fann 35 viscometer. A higher concentration (1.4 wt%)
xanthan (Table 2) was tested which also deviated from power-law
behavior. At longer times, the 1.4 wt% solution was dominated by low
shear rates (b0.1 1/s) where a low-shear Newtonian plateau exists.
4.3. Bentonite
Solutions with bentonite are often reported as exhibiting a yield
stress. Fig. 6 shows the Fann 35 rheology data for 8.6 wt% (30 lb/bbl)
bentonite; both a power-law and HerschelBulkley model ts the
data well. However, less subjectivity is observed in the Marsh Funnel
compared to the rheometers. The height versus time curve
approaches an asymptote (Fig. 7) and the ow is observed to come
to a complete stop at h=16 cm (not even a drop was observed for
several days), which corresponds to a yield stress of
0
=47 Pa. The
remaining HerschelBulkley parameters (m and n) were then found
from a best t to the funnel model.
The Fann 35 data show a minimum stress of 11 Pa suggesting that

0
11 Pa, contradicting the funnel data. A yield stress of 11 Pa is not
large enough to suspend uid in the funnel and the model predicts
(Fig. 8) that uid drains relatively quickly for the estimated Fann 35
parameters compared to the observed behavior. The Fann 35
Fig. 4. Rheological data for 1.1 wt% xanthan presented as (a) shear stress versus shear
rate and (b) viscosity versus shear rate on a loglog plot. Best t parameters n=0.10
and m=15 Pa-s
n
for the Fann 35 and n=0.09 and m=16 Pa-s
n
for the ARES were
found.
Fig. 5. Height versus time funnel data for 1.1 wt% xanthan. A best-t solution for the
model was found using
0
=220 Pa-s,

=0.014, m=18 Pa-s

n
, and n=0.12 were
found. The predicted curve using the Fann 35 estimated parameters is also included.
Table 2
Rheological parameters obtained via best t to various models (power-law, Carreau, or
HerschelBulkley).
Fluid Rheometer Shear rate
(1/s)
m
(Pa-s
n
)
n
0
(Pa-s)

(Pa-s)

0
(Pa)
Mineral oil Funnel 0-400 0.18 1 0
Fann 35 11000 0.19 1 0
ARES 0.01-3000 0.20 1 0
Xanthan
(1.1 wt%)
Funnel 03500 18 0.12 220 0.014 0
Fann 35 11000 15 0.10 0
ARES 0.01-3000 16 0.09 0
Xanthan
(1.4 wt%)
Funnel 01000 24 0.12 290 0.019 0
Fann 35 11000 15 0.14 0
ARES 0.01-3000 23 0.09 0
Bentonite
(7.2 wt%)
Funnel 0700 11 0.28 0
Fann 35 11000 8 0.28 0
Bentonite
(8.6 wt%)
Funnel 0200 0.76 0.64 47
Fann 35 11000 0.69 0.74 11
Carbopol
(2 wt%)
Funnel 0100 8.4 0.32 31
Fann 35 11000 2.1 0.63 32
HPAM
(1 wt%)
Funnel 0500 2.5 0.56 0
Fann 35 11000 7.0 0.27 0
ARES 0.01-3000 4.9 0.24 0
398 M.T. Balhoff et al. / Journal of Petroleum Science and Engineering 77 (2011) 393402
rheometer may underestimate the yield stress because the shear rate
(as opposed to shear stress) is imposed. As a result, the three-
dimensional microstructure of the uid may break even at low rates,
thus yielding the material. On the other hand, the stress is controlled
in the funnel by the uid height. The stress is slowly reduced and uid
comes to a complete stop. As a result, the three-dimensional
microstructure remains intact. After reaching a static height, the
bentonite uid was manually disturbed in the funnel; the micro-
structure broke and began to ow again until it reached a new static
height (not shown in the gure).
4.4. Carbopol
Experiments using 2 wt% and 3 wt% carbopol were conducted
since it was expected to exhibit a yield stress. 3 wt% carbopol did not
ow at all from the original height of 27.5 cm (not even a drop was
observed) indicating
0
N75 Pa. One limitation of the funnel is the
limited range of imposed stress in the upper cone (1275 Pa). This
problem could easily be addressed by using funnels of different size
and shape. 2 wt% carbopol was found to have a yield stress of 31 Pa
and the funnel ow behavior was similar to that of bentonite.
Estimated parameters are shown in Table 2.
4.5. HPAM
Figs. 8 and 9 show the rheometer and funnel data, respectively, for
1 wt% HPAM. The uid did not exhibit a yield stress and appeared to
be mostly in a shear-thinning regime; therefore a power-law model
seemed to be sufcient. Power-law parameters n=0.27 and
m=7.0 Pa-s
n
were found for the Fann 35 which matched relatively
well to the ARES (n=0.24 and m=4.9 Pa-s
n
). Although the funnel
data matched a power-law model very well, the best-t parameters
were very different from the rheometers' (n=0.56 and m=2.5 Pa-
s
n
); HPAM was the only uid which resulted in a signicantly
different shear-thinning index (n) for the funnel when compared to
the Fann 35.
Fig. 6. Rheological data for 8.6 wt% bentonite using the Fann 35 rheometer. Best-t
parameters, n=0.74, m=0.69 Pa-s
n
, and
0
=11 Pa, were found.
Fig. 7. Height versus time data for 8.6 wt% bentonite. Best-t parameters to the funnel
model were found as n=0.64, m=0.76 Pa-s
n
, and
0
=47 Pa. The best-t model is also
compared to the prediction based on the Fann 35 estimates.
Fig. 8. Rheological data for 1 wt% HPAM. Best t Fann 35 parameters n=0.27 and
m=7.0 Pa-s
n
for the Fann 35 and n=0.25 and m=4.9 Pa-s
n
for the ARES were found.
Fig. 9. Height versus time data for 1 wt% HPAM. A best-t solution for the model was
found using m=2.4 Pa-s
n
, and n=0.57.
399 M.T. Balhoff et al. / Journal of Petroleum Science and Engineering 77 (2011) 393402
HPAM is well-known to be viscoelastic and exhibit hysteresis and
other time-dependent effects (Lake, 1989; Carreau et al., 1997;
Delshad et al., 2008). The funnel model does not currently account for
any of these phenomena and ow through the constriction from the
upper cone to the capillary tube may be largely inuenced by
elasticity. The elastic uid elongates as it passes through the
constriction and then relaxes in the tube. Moreover, the stress
response to a change in shear rate is not instantaneous and rheological
measurements should be taken at steady state. In the funnel, a steady
state shear rate is never reached (it decreases as uid drains) which
could lead to inaccurate predictions of steady-shear behavior. For
uids with minimal elastic effects, the slow change of shear rates may
not signicantly affect results, but these effects may partially explain
the discrepancy between the rheometers and funnel results for HPAM.
The model could be potentially improved by including an elastic
component in the model. Several constitutive equations for stress
(Maxwell, Oldroyd-B, etc.) exist, but derivation of resulting ow
equations through a constriction can be difcult. Future work might
focus on computational uid dynamics modeling of viscoelastic ow
in the funnel to investigate its effect on drainage.
5. Discussion
Table 2 summarizes the estimated rheological properties for all
uids tested assuming a shear-thinning (Power-Law or Carreau) or
yield stress (HerschelBulkley) model. In most cases, very good
agreement occurs between the Fann 35 and funnel rheometers for
shear-thinning uids (elastic HPAM being the exception), despite
several assumptions. The shear-thinning index (n) was very consis-
tent between the Fann 35 and funnel for most uids. The consistency
index (m) was comparable, but higher in the funnel (~30%) on
average. The additional resistance could be due to the many model
assumptions (resistance in the tube alone, no entrance or exit effects,
and elasticity ignored). The funnel method has some advantages over
other portable lab rheometers (e.g. Fann 35) of more data points,
ability to calculate parameters describing Newtonian plateaus at low
and high shear rates, and a more objective measurement of uid yield
stress.
Table 2 also shows reasonable agreement for yield stress uids
(8.6 wt% bentonite and 2 wt% carbopol). Although yield stress
measurements are notoriously difcult to obtain, the funnel provides
a simple and objective measurement of yield stress. Unlike extrap-
olation techniques (on curves of stress versus shear rate) or even
static tests (e.g. Creep), there is little ambiguity in measurement. The
table shows that the funnel estimated a higher yield stress than the
Fann 35 viscometer for bentonite, possibly because the three-
dimensional microstructure broke in the controlled-rate viscometer.
The height versus time data necessary for solution to the model is
easy to obtain and, in theory, could be obtained with a beaker and
stopwatch. Solution to the ODE is robust and convergence to a
minimum sum of squared residuals for parameter estimation is good
provided a decent initial guess is given (e.g. the Fann 35 values).
However, alternatives to the solution of the ODE for data reduction are
possible. For example, plots of owrate (dV/dt) versus hydrostatic
head data could be generated and the ow equations in Table 1 t to
the data. Equivalently, the data could be converted to curves of
apparent viscosity versus shear rate. Data reduction might be even
easier using these approaches, but requires differentiating (instead of
integrating) the data and more scatter is observed (especially near the
end of the experiment when Q approaches zero).
Another alternative to the approach presented in this work is to
use only a few data points of height to estimate parameters. In most
cases the model t of height versus time was excellent suggesting that
only two data points would be necessary to obtain parameters for a
power-law uid. Two experiments could be conducted with different
initial volumes (e.g. 1000 mL and 1500 mL) and the power-law
parameters extracted explicitly from only the total drainage time of
the experiments. The ODE (Eq. (4)) would still need to be solved, but
it eliminates the need to obtain entire curves of height versus time.
6. Conclusions
We present a new method for obtaining rheological properties of
drilling muds and other non-Newtonian uids that uses height versus
time data in a draining Marsh Funnel. Currently the Marsh Funnel is
used in the oileld industry to estimate a relative viscosity of drilling
uids, but current practice uses only one data pointthe total
drainage time. Consequently, the procedure can be used to unambig-
uously estimate only one rheological parameter. The new method can
be used to estimate several parameters, including the uid yield
stress. The method is simple and very inexpensive; in theory it can be
performed with a funnel, beaker, and a stopwatch. The new funnel
approach appears to be as accurate as the popular Fann 35 viscometer.
The Marsh Funnel could be used as a tool in the eld for areas in the oil
industry other than drilling (e.g. enhanced oil recovery and
fracturing). A few major conclusions of this work are as follows.
1. There are excellent ts between the height versus time data and
the model (typically the curves lie on top of each other). This suggests
that a full data set used for least-squares t may not be necessary.
Instead for a two-parameter model, drainage time could be measured
for two different initial heights and the parameters calculated explicitly.
2. Fluids that have height versus time data that do not match the
theoretical model well after a least-squares t, probably should be
described by a more complicated rheological model. For example, the
funnel shear rate may be in a Newtonian plateau, which means that a
better constitutive equation (e.g. 4-parameter Carreau model) is
needed. Additionally, some uids exhibit strong elastic effects (e.g.
HPAM) and additional work should be performed to account for the
behavior in the model.
3. The ultimate static height in the funnel is a method to
objectively estimate yield stress. High-concentration solutions of
bentonite and carbopol drained to a static height from which
calculation of uid yield stress is simple and unambiguous. There is
no extrapolation. This approach has several advantages over other
existing methods that are subjective and/or difcult to perform.
7. List of variables
g gravity (cm/s
2
)
h height of uid in the funnel (cm)
h
0
initial height in the funnel (cm)
h
ss
steady state height in the funnel (cm)
H
F
total height of the cone portion of the funnel (cm)
L length of the capillary tube (cm)
m consistency index (Pa-s
n
)
n shear-thinning index
P pressure (Pa)
Q ow rate (cm
3
/s)
r radius of uid in the funnel (cm)
R radius of capillary tube (cm)
R
F
maximum radius of the funnel (cm)
t time (s)
t
f
total drainage time (s)
V volume of uid in the funnel (cm
3
)
coefcient relating volume and height
shear rate (1/s)

w
shear rate at the wall of capillary tube (1/s)
non-Newtonian viscosity (Pa-s)

0
low shear rate viscosity (Pa-s)

high shear rate viscosity (Pa-s)

400 M.T. Balhoff et al. / Journal of Petroleum Science and Engineering 77 (2011) 393402
Carreau model parameter (s)
viscosity (Pa-s)
uid density (g/cm
3
)

w
shear stress at the wall of capillary tube (Pa)

0
yield stress (Pa)
Acknowledgements
We would like to thank the Petroleum and Geosystems Depart-
ment and UT-Austin as well as the Yates Foundation for funding this
project. We would also like to thank Chun Huh and Do Hoon Kim for
the many thoughtful discussions on the subject. Larry W. Lake holds
the W.A. (Monty) Moncrief Centennial Chair.
Appendix. Verication of assumptions in funnel model
The mathematical model for uid height in the funnel was derived
in Eqs. (1)(10) assuming that the all of the resistance for a owing
uid is in the lower tube and none in the upper cone. Additionally, it is
assumed that ow is quasi-static so the steady-state momentum
equations can be utilized.
Newtonian uids without a yield stress
The total pressure drop in the funnel is the sum of pressure drops
in series of the tube and upper cone:
P
total
= P
tube
+ P
cone
A 1
For a Newtonian uid, the pressure drop in the lower tube is
Poiseuille ow,
P
tube
=
8L
R
4
tube
Q A 2
In the upper cone, ow is not unidirectional but a lubrication
approximation can be used to give a good approximation:
P
funnel
h
=
8
r h
4
funnel
Q A 3
Where the cone radius increases linearly from the connection of
the upper cone and capillary tube.
r h = R +
R
F
R
H
F
h
R
F
H
F
h A 4
Substituting and separating variables gives
P =
8

H
F
R
F
_ _
4
Q
H
F
0
h
4
dh A 5
And solving the resulting ODE, gives:
P
cone
=
8

H
F
R
F
_ _
4
Q
1
3H
3
F
A 6
The ratio of pressure drops is now:
P
cone
P
tube
=
R
tube
R
F
_ _
4
H
F
3L
= 2 10
6
bb1 A 7
This calculation is based on the height of the top of funnel which
represents the most possible resistance of the cone. The pressure drop
because of viscous ow in the upper cone is negligible for Newtonian
uids. This is mainly because the radius quickly becomes large and the
resistance is inversely proportional to r
4
. Resistance in the upper cone
is even smaller for shear thinning uids without a yield stress because
resistance is inversely proportional to r
3+1/n
. Since n is less than one,
the pressure drop because of viscous ow is even less signicant.
Yield-stress uids
For yield stress uids there is additional pressure loss (outside of the
viscous forces derived in Eq. (A-7) in the upper cone). At steady state
(h=h
ss
) there is no owin the funnel (Q=0) and Eq. (A-1) can then be
written as:
P
total
=
2L
0
R
+
2h
0
r h
funnel
A 8
Eq. (A-4) gives:
P
total
=
2L
R
+
2H
F
R
F
_ _

0
A 9
This leads to the following expression for yield stress

0
=
g h
ss
+ L
2L
R
+
2H
F
R
F
_ _ A 10
This also implies that the total pressure drop should not be used for
the Bingham and HerschelBulkley ow equations (Table 1). For
relatively large ow rates the pressure drop in the cone is small but
becomes signicant as the drainage rate decreases and yield stress
effects become important. The pressure loss in the tube should be
corrected as follows:
P
tube
= g h + L
2H
F
R
F

0
A 11
Quasi-steady ow assumption
Another assumption included in the model is quasi-steady ow;
that is time-dependent effects associated with the uid head are
neglected. A transient momentum balance describing ow in the
capillary tube is given as:
v
z

t
= g
z
+
1
r

r
r
rz

_ _
A 12
The assumption is that the time derivative is small compared to
the other two terms on the right hand side (gravity and viscous
resistance). For constant density uids, this requires:
v
z

t
bbg
z
A 13
The assumption is valid for all uids tested. For all experiments
conducted in this work, the unsteady term is factors of ten smaller
than gravity. For example mineral oil has one the largest time
derivatives and it is on average only 210
5
g
z.
References
Balhoff, M., 2005. Modeling the Flow of non-Newtonian uids at the Pore Scale. PhD
Dissertation. Louisiana State University.
Balhoff, M.T., Miller, M.J., 2005. An analytical model for cleanup of yield-stress uids in
hydraulic fractures. Soc. Pet. Eng. J. 10 (1), 311.
Barnes, H., Walters, K., 1985. The yield stress myth? Rheol Acta 24, 323326.
Bourgoyne Jr., A.T., Chenevert, M.E., Millheim, K.K., Young Jr., F.S., 1991. Applied Drilling
Engineering, SPE Textbook Series, Vol. 2. Society of Petroleum Engineers,
Richardson, TX.
401 M.T. Balhoff et al. / Journal of Petroleum Science and Engineering 77 (2011) 393402
Carreau, P.J., De Kee, D.C.R., Chhabra, R.P., 1997. Rheology of polymeric systems. Hanser/
Gardner Publications, Inc., Cincinatti.
Chase, G.G., Dachavijit, P., 2003. Incompressible cake ltration of a yield stress uid.
Sep. Sci. Tech. 38 (4), 745766.
Delshad, M., Kim, D.H., Magbagbeola, O.A., Huh, C., Pope, G.A., 2008. Mechanistic
interpretation and utilization of viscoelastic behavior of polymer solutions for
improved polymer-ood efciency. SPE 11620 presented at the 2008 Improved Oil
Recover Symposium, Society of Petroleum Engineers, Tulsa, Oklahoma.
Fann 35 Instruction Manual, 2010. http://www.Fann.com Last accessed August 6, 2010.
Gray, G.R., Darley, H.C.H., Rogers, W.F., 1980. Composition and Properties of Oil Well
Drilling Fluids. Gulf Publishing Company, Houston, pp. 181276.
Lake, Larry W., 1989. Enhanced Oil Recovery. Prentice Hall, NewJersey, pp. 216, 4392,
317353. Available through author.
Marsh, H., 1931. Properties and Treatment of Rotary Mud. Petroleum Development and
Technology, Transactions of the AIME, pp. 234251.
May, E.A., Britt, L.K., Nolte, K.G., 1997. The Effect of Yield Stress on Fracture Fluid
Cleanup. SPE 38619, presented at the 1997 Society of Petroleum Engineers Annual
Technical Conference and Exhibition in San Antonio, Texas.
Moller, P., Fall, A., Bonn, D., 2009. Origin of apparent viscosity in yield stress uids below
yielding. EPL 87 (3), 16.
Nguyen, Q.D., Boger, D.V., 1983. Direct yield stress measurement with the vane method.
J. Rheol. 29 (3), 335347.
Nguyen, Q., Boger, D., 1992. Measuring the ow properties of yield stress uids. Annu.
Rev. Fluid Mech. 24, 4788.
Peder, C., Moller, P., Mewis, J., Bonn, D., 2006. Yield stress and thixotropy: on the
difculty of measuring yield stresses in practice. Soft Matter. 2, 274283.
Pitt, M.J., 2000. The Marsh Funnel and drilling uid viscosity: a new equation for eld
use. Soc. Petroleum Eng., Drilling Completions 15 (1), 36.
Skelland, A.H.P., 1967. Non-Newtonian ow and heat transfer. John Wiley & Sons, New
York.
Song, K., Kim, Y., Chang, G., 2006. Rheology of concentrated xanthan gum solutions:
steady shear ow behavior. Fibers Polymers 7 (2), 129138.
Zhu, L., Sun, N., Papadopoulos, K., De Kee, D., 2001. A slotted plate device for measuring
yield stress. J. Rheol. 45 (5), 11051122.
402 M.T. Balhoff et al. / Journal of Petroleum Science and Engineering 77 (2011) 393402