ANNUAL TRANSACTIONS OF THE NORDIC RHEOLOGY SOCIETY, VOL.

16, 2008

Thixotropy in Water-Based Drilling Fluids

Ahmadi Tehrani

M-I SWACO Research and Technology Centre, Aberdeen, United Kingdom

ABSTRACT approximation, the time- and shear-

Rheology of drilling fluids affects the dependence of rheological parameters
frictional pressure drop and the solids within the range of shear rates encountered
carrying capacity of the fluids during the in the drillstring and in the annulus between
drilling operation. Drilling fluid rheology is the drillstring and the borehole wall. The
commonly controlled by using a variety of model shows potential for use in calculation
clay or polymeric materials, depending on of drilling hydraulics.
the type of fluid used and the demands of
the specific drilling operation. Most drilling INTRODUCTION
fluids possess varying degrees of time- and The function of a drilling fluid is to cool
shear-dependent thixotropic properties. and lubricate the drill bit, transport cuttings
Among them, water-based fluids containing to the surface, and stabilise the wellbore.
clays, such as bentonite, exhibit a For the fluid to perform satisfactorily,
pronounced thixotropic behaviour. This frictional pressure drop and solids-bearing
characteristic can have a significant effect capacity must be maintained at an optimum
on the peaks and troughs of pressure that level throughout the drilling operation. The
occur in the wellbore when the drillstring or rheological parameters that control these
the tool-string is moved in and out, or when properties are the fluid viscosity and the
pumping starts after a break in circulation. yield stress. Efficient pumping requires a
Major pressure fluctuations can lead to low enough viscosity while adequate yield
fracturing of the formation, loss of stress is needed to maintain cuttings in
circulation, influx of formation fluids into suspension, particularly during circulation
the wellbore, or collapse of the wellbore. breaks.
Thus, a practical means of accounting for Drilling fluid rheology is commonly
thixotropy in hydraulics calculations would controlled by using a variety of clay or
be of great value for maintaining safety as polymeric materials, depending on the type
well as the integrity of the wellbore. of fluid used and the demands of the specific
As far as can be determined the drilling operation. Of the various types of
thixotropic characteristics of drilling fluids drilling fluids, water-based fluids (also
have not been quantified for engineering referred to as water-based mud, WBM)
applications. In this work, a simple model containing clays such as bentonite (or
based on the concept of a structure particulates such as mixed metal oxides or
parameter is used to describe drilling fluid hydroxides) exhibit a pronounced
thixotropy. Empirical relationships are thixotropic behaviour. The clay-based fluids
devised that can predict, with good are suspensions of bentonite in water to
which various compounds such as heavier described by a double-exponential function.
minerals, polymers and surfactants are More recently, Dolz et al.7 investigated the
added to control density, rheology and fluid thixotropic behaviour of water-based
loss, and to improve other fluid properties drilling fluids containing bentonite, at 6-
such as shale inhibition and lubricity. The 12% (w/w) concentration, and a polymeric
suspended clay particles are thin, flat material. They found that the lower
platelets that are electrically charged and concentrations of bentonite produced the
interact to form a loose house-of-cards greatest thixotropic effect. They obtained an
structure that is responsible for the gelling empirical equation that relates shear stress
characteristics of the fluid when at rest, and to the concentrations of the thickeners, the
its thinning behaviour when sheared.1 This shear rate and the formulation stirring time.
structure and the resulting bulk rheological As far as can be determined, the
properties are time- and shear-history thixotropic characteristics of drilling fluids
dependent and the fluid is said to possess have not been quantified for engineering
thixotropy properties. applications. To improve the quality of
The rheological properties of the drilling downhole predictions, there is a need for
fluid are subject to continuous modification some form of engineering relationship that
as the fluid travels around the wellbore. This can describe the time- and shear-history
includes changes caused by variations in dependence of rheological properties in
shear rate, temperature, pressure, as well as drilling fluids. Currently, some of the more
chemical modification of the fluid as it advanced hydraulics modelling calculations
contacts various formations on its way to the account for this effect by using the so-called
surface. In addition, mechanical and thermal “gel” values that are measured by the API-
stresses that lead to the degradation of the standard oilfield viscometric method. In this
drilling fluid additives can cause significant method, the shear stress of the fluid is
changes in fluid rheology.2 measured after a 10-second, 10-minute, and
The shear rates to which the fluid is sometimes 30-minute rest period following
subjected can range from around 103 s-1 in a short interval of shearing at a high rate. A
the drillstring (as the mud travels down the requirement of this approach has been for
well), to ~105 s-1 in the highly turbulent flow the design of fluids with non-progressive
as the fluid issues from the drill bit, to those gels, i.e. where the longer-term gels are not
prevalent in the annulus, i.e. 0-102 s-1 significantly higher than the 10- or 30-
(depending on the eccentricity of the minute gel values. This approach has proved
annular space), as it carries the drill cuttings to be adequate for the applications
to the surface. encountered to date. However, as drilling
The combination of temperature, scenarios become more complex, a need for
pressure, composition, time- and shear- a better way of accounting for thixotropy of
history dependence of the bulk rheological drilling fluids will inevitably arise. An
properties makes a full characterisation of example of this is in depleted-zone drilling
the drilling fluid rheology a complex task. A where the operating window for fluid
number of works have been reported on the density (mud weight), between the
temperature, pressure and composition maximum and minimum dictated by the
dependence of water-based drilling fluids,2-5 pore pressure and fracture gradient
but only few references exist in the literature considerations, becomes very narrow. In
on attempts to quantify the thixotropic such circumstances, more precise
behaviour of such fluids. Mercer and knowledge of fluid thixotropy may help to
Weymann6 investigated the time better control drilling hydraulics.
dependence of viscosity in bentonite-water The work reported here uses a simple
suspensions and observed that it can be model based on the concept of a structure
parameter to describe drilling fluid shear stress, Fig. 2. The breakdown of
thixotropy. Empirical relationships are structure, which occurs as shear rate is
devised that can predict, with good increased along the “up” curve, is not fully
approximation, the time- and shear- recovered during the “down” curve and the
dependence of rheological parameters material completes the cycle with some
within the range of shear rates encountered residual broken structure. The area enclosed
in the drillstring and the annulus. The model between the “up” and “down” curves (the
requires further development but shows hysteresis loop) is an indication of the extent
potential for use in drilling hydraulics of thixotropy of the material.
calculations.

THIXOTROPIC FLOW BEHAVIOUR 140 Shear Rate = 1460

Shear Rate = 1460
Shear Rate = 461
Shear Rate = 146.1
Thixotropic materials are fluids 120 Shear Rate = 1460 Shear Rate = 46

Shear Stress (Pa)

Shear Rate = 1460 Shear Rate = 11.6
containing some form of structure as a result 100
Shear Rate = 1460
of formation of flocs or aggregates between 80
suspended particles or moieties. In clay 60
suspensions the formation of structure is 40
promoted by increased encounter between 20
suspended particles, which can result from 0
Brownian motion of the particles or from 0.0 10.0 20.0 30.0 40.0
the velocity gradient when the bulk of the Time (sec/103)

material is sheared. Structure breakdown Figure 1. Shear rate step-change tests.

can be due to collision of particles and flocs
as well as viscous drag exerted by the liquid 120
medium when the material is sheared. On a
100
smaller scale, the Brownian motion of
Shear Stress (Pa)

primary particles making up a floc can also 80

cause floc breakup. This means that under 60

certain conditions both Brownian motion 40
and shear can cause structure breakdown.
20
Thixotropic behaviour occurs when the
buildup effect of Brownian motion is 0
0 300 600 900 1200 1500
dominant over the breakdown effect of -1
Shear Rate (s )
shear.
When a thixotropic material is sheared, Figure 2. Hysteresis loops demonstrating
the buildup and breakdown processes thixotropic behaviour.
compete and a dynamic equilibrium
eventually results. Since the rates of buildup If the material is now left to rest, the
and breakdown of structure are finite, if broken structure will gradually re-form.
conditions are displaced from equilibrium However, if it is subjected to successive
(eg. by a change in shear rate), the structure ramping cycles until two consecutive loops
level will take some time to adjust. The superpose, then there is no further drop in
change in structure will be detected by a the level of structure and the loop is called
corresponding change in shear stress. This is the equilibrium loop. Another characteristic
illustrated in the shear rate step-change of such rheograms is the equilibrium flow
experiments of Fig. 1. curve. This is obtained by allowing the
Thixotropic behaviour may also be material to reach equilibrium structure level
observed by ramping the shear rate up or at each new shear rate such that the “up”
down and recording the resulting changes in
and “down” curves superpose and there is 1
no hysteresis. λe = (4)
b
The rate and extent of these changes are 1 + γ&
a
the parameters that describe thixotropic
behaviour and which can be determined by Moore also suggested an equation of
appropriate rheological experiments. state of the form:
Generalised Structure Theory
τ (t ) = [η ∞ + cλ (t )] γ& (5)
In order to obtain a qualitative
description of thixotropic behaviour above
yield point, a theory involving a single where η∞ is the viscosity at infinite shear
structure parameter λ has been developed rate and c is the incremental contribution of
which makes use of the concept of some structure to viscosity. Eq. 5 was later
rheological structure at the molecular and modified by Cheng and Evans 8 who
particulate level.8,9 The time dependence of allowed for the presence of a yield stress
rheological properties in a thixotropic proportional to the structure parameter λ(t):
material is then considered to be due to
changes in this structure. Two constitutive τ (t ) = λ (t )τ y + [η∞ + cλ (t )] γ& (6)
equations are used to describe single-
structure thixotropic behaviour; a rate In the case of water-based drilling fluids,
equation describing structure breakdown there is evidence (Alderman, et al.3, Zamora
and buildup as a function of shear rate and and Lord 10) that the equilibrium flow curve
λ: is satisfactorily represented by the Herschel-
Bulkley equation11:
dλ
= g (γ&, λ ) (1)
dt τ = τ H + kγ& n (7)

and an equation of state relating shear stress Incorporating the effect of structure on
to shear rate and structure parameter λ: yield stress and viscosity in Eq. 7 gives a
modified version of the Moore-Cheng
τ = τ (γ&, λ ) (2) equation (Eq. 5):

The simplest form of rate equation for τ (t ) = λ (t )τ y + [η ∞ + cλ (t )] γ& m (8)

structure breakdown and buildup at constant
shear rate is given by Moore: 9 Accordingly, the equilibrium flow curve
for the modified Moore-Cheng equation
dλ (t ) becomes:
= a[1 − λ (t )] − bλ (t )γ& (3)
dt 1424 3 123
buildup breakdown
τ e = λeτ y + (η ∞ + cλe ) γ& m (9)
in which it is assumed that the rate of
structure breakdown is proportional to both Equations 3 and 9 may be used as the
λ and shear rate, whereas the rate of buildup basis for a mathematical description of
is proportional to the broken structure (1-λ) thixotropic behaviour in fluids. The various
but is independent of shear rate. This rate parameters in Eq. 9 can be determined by
equation predicts a dynamic equilibrium measuring the steady-state shear stress at a
when the structure parameter is: number of shear rates. A plot of τe vs. shear
rate will then give values for parameters τy,
η∞, c and m in Eq. 9, provided that λe can be properties of the mud to any measurable
expressed as a function of shear rate by a extent.
relationship such as Eq. 4. Four types of measurements were
performed:
EXPERIMENTAL 1) Stress relaxation experiments: These
were carried out to determine the
Materials equilibrium stress at a number of shear
The material used in these experiments rates. The sample was sheared at a
was an unweighted WBM containing 64.3 g constant rate for a certain length of time.
bentonite per litre of water (supplied by M-I Some early experiments were allowed to
SWACO). This was prepared by adding the proceed for up to 25 hr. The results from
clay to water over a period of two minutes these showed that the stress values
while agitating the suspension with a became steady after 3-5 hr. Thereafter,
Silverson high-speed mixer. The pH of the most of the experiments were of 6-8 hr
suspension was measured to be about 9.5. duration. The experiments were repeated
The suspension was hot-rolled in an API 3-6 times at each shear rate and showed
standard roller oven at 100˚C and at an good reproducibility. The shear rates
approximate shear rate of 5 s-1, for a period used were in the range 2.9–1460 s-1.
of 48 hr. 2) Shear rate step-change experiments: In
these measurements the shear rate was
Experimental Procedure
held constant at an initial value γ&1 until
Most of the experiments were carried out
on a Bohlin VOR rheometer, using the C25 the shear stress reached a steady final
concentric cylinder geometry as the value. Then the shear rate was abruptly
measuring system. The radii of the bob and changed to a new value γ&2 and the time-
cup of the geometry were 12.5 and 13.75 dependent shear stress was measured to
mm, respectively, and the bob had a height equilibrium. This procedure was
of 37.5 mm. The Carrimed controlled-stress repeated several times between four
rheometer was used for yield-stress different pairs of high-low or low-high
measurements with concentric cylinder shear rates. The results of these
geometry. Prior to each experiment, the experiments were used to establish the
material was cold-rolled for one hr at room time- and shear-history dependence of
temperature and at a shear rate of 5 s-1. This shear stress.
had a homogenising effect on the fluid and 3) Yield stress measurements: These were
brought all samples to similar initial shear carried out on the controlled-stress
history. The samples were allowed to rest in rheometer. Shear stress was applied to
the measuring system of the rheometer for the sample and increased at a controlled
ten minutes before measurements began. rate until angular displacement was
The sample temperature was maintained at detected.
25 ±0.1°C throughout each experiment.
To minimise sample dehydration during 4) Hysteresis loops: In these experiments
measurements, a thin layer of a viscosity the shear rate was ramped repeatedly at
standard oil S60 (supplied by Cannon a controlled rate over a selected range
Instruments Company, USA), with a and the corresponding shear stress was
viscosity of 102.3 mPa·s at 25°C, was measured. The aim of these experiments
placed on the free surface of the sample. A was to obtain the hysteresis loops
series of early experiments confirmed that characteristic of thixotropic materials.
the oil film did not affect the rheological
RESULTS AND DISCUSSION 100
Fig. 3 shows typical stress-relaxation
data at several shear rates. The results 80

Shear Stress (Pa)

indicate that there is a relatively sharp drop 60
in shear stress in the early stages of
shearing, consistent with a fast initial 40

structure breakdown. 20

0
120 0 300 600 900 1200 1500
Shear Rate = 1460 Shear Rate = 1160 -1
100 Shear Rate (s )
Shear Rate = 461 Shear Rate = 146.1
Shear Stress (Pa)

80 Shear Rate = 91.9 Shear Rate = 23.2 Figure 4. Equilibrium flow curves.
60
120
40 Shear Rate = 461 Shear Rate = 91.9
100
20 Shear Rate = 461 Shear Rate = 91.9

Shear Stress (Pa)

80 Shear Rate = 461
0
0.0 5.0 10.0 15.0 20.0 25.0 60
Time (sec/103)
40
Figure 3. Steady-shear stress-relaxation
20
data.
0
0.0 10.0 20.0 30.0 40.0
In most of the experiments, the shear Time (sec/103)
stress reached a steady value τe within the
duration of the experiment, but at the lower Figure 5. Shear rate step-change tests.
shear rates (<100 s-1) shear stress continued
to decline even after 8 hr, suggesting long Single-Structure Model
relaxation times. Subsequently, some of To find the parameters in the modified
these experiments were repeated during the Moore-Cheng equation (Eq. 8), it is
shear rate step-change measurements such necessary to express λe as a function of
that equilibrium was approached both from shear rate. Eq. 4 is one such relationship
above (net breakdown) and from below (net based on the single-structure theory by
buildup) in order to find the true τe. Moore (Eq. 3). This theory also yields a
Fig. 4 is a plot of the equilibrium stress single-exponential time dependence for λ (at
τe as a function of shear rate. The data constant shear rate), obtained by integrating
indicate that the material has a finite yield Eq. 3,
stress and that the change in the slope of the 1
flow curve decreases as the shear rate is λ (t ) = (λ0 − λe ) e
−( )
T
+ λe (10)
increased. This is characteristic of the
Herschel-Bulkley type fluids.
An example of the shear rate step- where,
change experiments is illustrated in Fig. 5,
where stress relaxation is in the form of 1
= a + bγ& (11)
decay or recovery depending on whether the T
shear rate is increased or decreased.
and λ0 is the level of structure at time t = 0.
Eq. 10 can be used to define the time
dependence of shear stress when substituted the lower shear rates to 0.02 at the highest
in Eq. 8: shear rate, another indication that the initial
breakdown, which is faster at higher shear
1
−( ) rates, is not well predicted.
τ (t ) = τ 1 e T
+τ e (12) Eq. 12 gives a better average fit (χ2 = 7.0
× 10-5) to the shear stress-time data obtained
where, in the rate step-change experiments, because
here the material has already been sheared
τ 1 = (λ0 − λe )(τ y + cγ& m ) (13) and has a lower level of structure.
Consequently, the initial breakdown by
viscous drag is less severe, and the single-
τ e = λeτ y + (η ∞ + cλe ) γ& m (14) exponential function gives a closer
prediction.
The validity of the single-structure rate Fig. 6 is a plot of the pre-exponential
equation for the WBM of these experiments coefficient τ1 vs. γ& (Eq. 12). The data
was tested by fitting a 3-parameter single-
exhibit a steep rise with γ& at lower shear
exponential function [τ1, T and τe in Eq. 12]
to the experimental shear stress-time data rates. A similar behaviour is observed in the
and using a non-analytical least-squares plot of T-1 vs. γ& (Eq. 11), Fig. 7.
regression routine. These are shown as the
solid lines in Fig. 3 for some of the stress-
40
relaxation experiments. Although the
general fit is good, it seems that such
Shear Stress (Pa)

30
function does not describe the data
satisfactorily at early times in the 20
experiments. This might suggest that in the
stress-relaxation experiments, where the 10
material is sheared from rest, i.e. a state of
relatively high structure, there is a fast 0
viscous drag-driven structure breakdown (to 0 300 600 900 1200 1500

flocs and aggregates) followed by a slower Shear Rate (s-1)

breakdown caused by the collision of mobile Figure 6. Pre-exponential coefficients from

flocs and aggregates. Because the a single-exponential fit to stress-relaxation
population of data points is heavier toward data.
the steady state, it is mostly the behaviour of
the slower process that is reflected in the
3.0
fitted function.
At constant shear rate, the regression 2.5

coefficient χ2 may be defined as: 2.0

T (10 s )
-1
-3

1.5
N
1
χ =
2
∑ [τ (t ) − τ (ti ) calc. ] 2
-1

1.0
2 i meas .
( N − M )τ e i =1
0.5
(15) 0.0
0 300 600 900 1200 1500
in which N is the number of data points and -1
Shear Rate (s )
M is the number of parameters in the
function to be fitted. For the stress Figure 7. Inverse relaxation time vs. shear
relaxation data, χ2 ranged from 2.5 × 10-5 at rate for stress-relaxation experiments.
Time-Dependence of Shear Stress Table 1. Relaxation times of structure
According to the single-structure rate breakdown and buildup processes.
theory, the relaxation time decreases with Shear rate
To
increasing shear rate (Eq. 11); a plot of T-1 (s-1)
vs. γ& is a straight line with slope b, and From 91.9 146.1 461.0 1460.0
intercept a on the T-1 axis. However, it 91.9 - - 608 143
seems from Fig. 7 that at lower shear rates,
up to about 100 s-1, the relaxation time 146.1 - - 661 331
decreases sharply with increasing γ& and
461.0 1550 1060 - 671
then follows a slower rate of decrease as
shear rate increases. 1460.0 1060 1230 910 -
A method advocated by many for
determining the time dependence of shear The results suggest that the buildup of
stress is the high-low shear rate step-change structure is a slower process than the
procedure illustrated in Fig. 5. Assuming breakdown. In one case (91.9 and 146.1 s-1)
single-exponential time dependence, a least- the close proximity of the two shear rates
squares regression of data at two shear rates made calculation of their relaxation times
gives the time constants for structure subject to large errors.
breakdown and buildup, T1 and T2, For each pair of relaxation times, Eqs. 16
respectively: and 17 were solved simultaneously to find
the constants a and b. These are given in
1 Table 2 for five pairs of the shear rates used.
= a + b γ&1 (16)
T1 Due to the wide variation in the b/a ratio, it
is unlikely that the time dependence of the
1 WBM can be adequately described by a
= a + b γ&2 (17) single relationship over the entire range of
T2 shear rates. A better approximation may be
obtained by fitting two straight lines to the
from which the constants a and b are T-1 vs. γ& data:
calculated. The range of shear rates used in
the step-change experiments was 100–1500 For shear rates in the range 0 – 100 s-1,
s-1. The results of these experiments are
presented in Table 1 as the relaxation times 1
= 1.5 ×10 −5 + 4.8 ×10 −6 γ& (18)
of the breakdown (low-to-high shear rate) T
and buildup (high-to-low shear rate)
processes. The average value of χ2 for these and for 100 – 1500 s-1,
calculations was 7.0 × 10-5, and the error in
the reported relaxation times was less than 1
= 1.5 ×10 − 4 + 7.0 ×10 −7 γ& (19)
1%. The relaxation times to the left of the T
diagonal belong to net buildup of structure,
while those to the right of the diagonal The straight lines in Fig. 7 show a
represent net breakdown. graphical representation of the above
relationships.
 Table 2. Rate constants for structure
breakdown and buildup processes. Differentiating Eq. 10 in a similar manner
Shear Rates a b/a gives:
b
(s-1) (s-1) (s)
dλ (t ) λ0 − λe −( Tt )
91.9 ↔ 461.0 4.0×10 -4
2.7×10 -6
0.0068 = e (23)
dt T
91.9 ↔ 1460.0 6.5×10-4 4.3×10-6 0.0066

146.1 ↔ 461.0 6.8×10-4 1.8×10-6 0.0026

Eliminating dλ(t)/dt between Eqs. 22 and 23
results in:
146.1 ↔ 1460.0 5.7×10-4 1.7×10-6 0.0029

9.2×10-4 0.4×10-6 τ1
461.0 ↔ 1460.0 0.0004 λ0 − λe = (24)
τ y + c γ& m
In the absence of a single description for
time dependence in the range of shear rates The τ1 vs. γ& data presented in Fig. 6 are
used here, the value of the b/a ratio will adequately described by a simple empirical
have to be determined by means of curve equation of the type:
fitting to the equilibrium stress data.
k1 γ&
Shear-History Dependence of Shear Stress τ1 = (25)
To describe structure as a function of 1 + k 2 γ&
time, it is necessary to find an expression for
λ0. The initial structure λ0 depends on the Combining Eqs. 24 and 25 yields,
previous shear rate(s) to which the material
may have been subjected. Therefore, λ0 is k1 γ&
λ0 − λe = (26)
the parameter that describes the shear- (1 + k 2 γ& ) (τ y + c γ& m )
history dependence of the rheological
properties. λ0 cannot be measured directly Parameters in Equilibrium Flow Curve
but it may be determined in terms of τ1, the The parameters τy, b/a, η∞, c and m in
pre-exponential coefficient in the time Eq. 9 have now to be determined by a
description of stress-relaxation curves, as combination of curve fitting (of the
described below. measured τe data to the modified Moore-
Differentiation of Eqs. 8 and 12 with Cheng model) and independent
respect to time, and at constant shear rate, measurements. The independent measure-
gives: ments are needed to reduce the number of
parameters determined by curve fitting so
dτ (t ) dλ (t ) that a unique “best fit” can be determined.
= (τ y + c γ& m ) (20)
dt dt

dτ (t ) τ −( t )
=− 1e T (21)
dt T

Eliminating dτ(t)/dt between Eqs. 20 and 21

leads to:
t
dλ (t ) τ1 −( )
= e T
(22)
dt T (τ y + c γ& m )
 Equilibrium Shear Stress (Pa) 100 35
High-Shear Asymptote 30
80

Shear Stress (Pa)

Modified Moore-Cheng Model 25
60 20

15
40
10
20
5

0 0
0 300 600 900 1200 1500 0 10 20 30 40
-1 -1
Shear Rate (s ) Shear Rate (s )

Figure 8. Illustration of the high-shear Figure 9. Determination of the yield stress

asymptote and the modified Moore-Cheng on the controlled-stress rheometer.
model.
Substituting for η∞ = 0.02 Pa·sm and τy =
One way of achieving this is by 16.0 Pa, the modified Moore-Cheng
assuming that the drilling fluid behaves like equation at steady state becomes:
a Bingham fluid at higher shear rates, i.e.
m~1, and that the τe vs. γ& data can be 16.0 c
τe = + (0.02 + ) γ& m (27)
approximated by a straight line in that b b
1 + γ& 1 + γ&
region (λ0 ~ 0). The slope of the high-shear a a
asymptote will then give η∞. Fitting a
straight line through the data above γ& = 500 In the above function, parameters c and
s-1 in Fig. 8 gives η∞ = 0.02 ±0.002. If it is b/a exhibit high sensitivity to small
assumed that this is a good estimate for η∞ variations in the value of m. For example, a
in the Moore-Cheng model, then it will have 1% change in m produces variations as high
the units Pa·sm. as 20% and 30% in c and b/a, respectively.
A second parameter to be determined As a result, the function still does not seem
independently is the yield stress τy. This was to give a single best fit. Since parameters c
achieved by using the Carrimed controlled- and b/a have similar mutual sensitivities, it
stress rheometer along with the concentric is necessary to force the value of m in Eq.
cylinder geometry. The test fluid was pre- 27.
treated in the same way as in all other In the family of the best fits obtained by
measurements, i.e. 10-min cold-rolling least-squares regression of the equilibrium
followed by a 10-min rest period to reach data, the value of m varies from 0.642 to
thermal equilibrium. The yield stress was 0.985. Assuming an average value of m =
determined by applying a known shear 0.81, Eq. 27 may be forced to produce
stress to the sample and observing the rate singular values for b/a and c:
of angular displacement. The shear stress
was increased gradually until a displacement b
= 5.44 ×10 −4 ± 9.46 ×10 −5 (28)
was detected. The measurements were a
repeated several times, and a reproducible
value of τy = 16.0 Pa was obtained (Fig. 9). c = 0.20 ± 0.015 (29)
This value was adopted as the yield stress
corresponding to the initial structural state The fitted curve represented by these
of the fluid as used in all of the experiments. parameters is illustrated in Figure 8.
The complete set of relationships
produced for describing the time- and shear-
history dependence of the rheological the fluid is subjected to a sequence of shear
properties of the WBM of these experiments rates γ&i , for a period ti at each shear rate, the
is as follows: structure λi at the end of interval i, becomes
the initial structure for the next interval at
τ (t ) = τ y λ (t ) + [η ∞ + c λ (t )] γ& m (30) shear rate γ&i +1 , i.e. λ0,i +1 = λi (ti ) .
The experimental data in Fig. 10 were
t
−( ) obtained by ramping the shear rate up and
λ (t ) = (λ0 − λe ) e T
+ λe (31)
then down over the range 11.6 – 1460 s-1.
The cycle was repeated five times. The ramp
where, time, i.e. total “up” or “down” time, was 48
min, equivalent to 20 readings at 144-
1 second intervals. The fluid was pre-treated
λe = (32)
b as described earlier. The model was used to
1 + γ&
a predict the level of structure at the end of
the time interval at each shear rate. This
k1 γ& level of structure then became λ0 for the next
λ0 − λe = (33) time interval. The same method was
(1 + k 2 γ& )(τ y + c γ& m )
followed for both “up” and “down” curves.

τ y = 16.0 (34) 100

η ∞ = 0.02 (35) 80
Shear Stress (Pa)

c = 0.20 (36) 60

m = 0.81 (37) 40
b
= 5.44 ×10 − 4 (38) 20
a
0
k1 = 0.145 (39) 0 300 600 900 1200 1500

k 2 = 0.00516 (42) -1
Shear rate (s )

and the time dependence is further defined Figure 10. Hysteresis loops produced by
by the following relationships: ramping the shear rate in the range 14.6-
1460 s-1.
For shear rates in the range 0 – 100 s-1:
The solid lines in Fig. 10 show the
1 calculated hysteresis loops. The model fails
= 1.5 ×10 −5 + 4.8 ×10 −6 γ& (43)
T to predict the first cycle satisfactorily. This
is expected because, as discussed earlier, the
and for 100 – 1500 s-1: single-exponential time dependence does
not fit the stress relaxation data well at early
1 times. However, as time progresses, i.e. as
= 1.5 ×10 − 4 + 7.0 ×10 −7 γ& (44) the cycle is repeated, the agreement between
T
the experimental and predicted values
Prediction of Hysteresis Loops improves. The fourth and fifth cycles
The validity of the model and the practically superpose, indicating that the
accuracy of the final curve fitting can now equilibrium loop (to be distinguished from
be tested by predicting the hysteresis loops the equilibrium flow curve) is obtained after
produced when the shear rate is ramped. If a few cycles. The maximum difference
between the measured stress values and breakdown and buildup are assumed to be
those predicted by the model for the final governed by two rate processes with
loop is about 7%. This value decreases as distinctly different relaxation times. In this
the ramp time is increased. For example, a scheme the first process may be a fast,
total ramp time of 90 min gave a maximum viscous drag-driven breakdown, which
difference of 5% in the last loop, while a breaks the intact (house-of-cards) structure
shorter time of 32 min increased the error to into flocs and aggregates. The second
9%. process has a slower but longer-term effect;
The slopes of the “up” and “down” it causes further breakdown by promoting
curves are determined to a large extent by collision between the mobile flocs and
the value of parameter m. It appears that the aggregates and modifies the orientation of
assumed average value of m = 0.81 produces the particles thus formed in order to fully
good agreement with the experimental data. accommodate the applied shear stress.
For the model to give a good prediction Within the remaining intact structure the
of thixotropic behaviour, one of the rate of collision and particle orientation is
following initial conditions must be negligible due to low mobility of the flocs
satisfied: and aggregates. Therefore, the two structural
1) Material initially sheared at a very high processes may be assumed to be of the
rate so that λ0 ~ 0. This corresponds to consecutive or series type.
the condition of the drilling fluid as it The observations made earlier with
issues at the drill bit, where it is regard to relaxation time-shear rate
subjected to shear rates on the order of relationship for the WBM of this study,
105 s-1. favour a two-structure model. In such a
2) Fluid at equilibrium at a finite shear rate model, the time dependence of structure at
constant shear rate may be described by a
such that λ0 = λe . This condition may
double-exponential function of the type:
correspond to the state of the fluid in the
circulation system at the surface. t t
3) Material initially at rest, as in the case −( ) −( )

when there is a break in fluid λ (t ) = λ1 e T1

+ λ2 e T2
+ λe (43)
circulation. In this situation, one may
assume that the initial conditions are the in which the coefficients λ1, λ2, λe, T1 and T2
same as those for the fluid in these are functions of shear rate. A similar
experiments, or if there has been for an function describing the time dependence of
extended break in circulation, one may shear stress at constant shear rate was fitted
assume that the fluid has a fully formed to the stress relaxation data of this study,
structure, i.e. λ0 = 1. and the values of χ2 were found to be better,
at least by a factor of two, than those
ALTERNATIVE THIXOTROPY MODELS obtained with the single-exponential
Due to its simplicity, the single-structure function. Mathematical derivations leading
theory does not consider the physical and to the time-description of structure have
chemical nature of the structure. A more been made for this model, and it appears
precise model for describing structural that the five coefficients involved in Eq. 43
processes under shear requires a good are complicated functions of shear rate
knowledge of the nature of structure as well which do not lend themselves to
as the type of changes that it undergoes determination by simple optimisation
when sheared. An alternative model that procedures. This makes the two-structure
may be considered in the absence of such model unsuitable for engineering
information is one in which structure applications.
 4. Annis, M.R. (1967), “High-Temperature
CONCLUSIONS Flow Properties of Water-Based Drilling
The thixotropic characteristics of some Fluids,” J. Pet. Tech., 19 (8), 1074.
drilling fluids may have a significant effect
5. Hiller, K.H. (1963), “Rheological
on their rheological properties in the time
Measurements on Clay Suspensions and
scale and range of shear rates encountered
Drilling Fluids at High Temperatures and
around the wellbore. Hitherto, the effect has
Pressures,” J. Pet. Tech., 15 (7), 779.
been accounted for in fluid design and
engineering calculations by the 10-sec and 6. Mercer, H.A. and Weymann, H.D.
10-min gel values measured routinely in (1974), “Structure of Thixotropic
drilling fluid laboratories. The results of this Suspensions in Shear Flow. III. Time-
work show that relatively simple semi- Dependent Behaviour,” Trans. Soc. Rheol.,
empirical models, based on a single- 82 (1), 199.
structure theory, may be devised in such a 7. Dolz, M., Jiménez, J., Jésus Hernández,
way that the effects of time and shear M., Delegido, J. and Casanovas, A. (2007),
history on rheology can be incorporated in “Flow and Thixotropy of non-
engineering calculations. Contaminating Oil Drilling Fluids
Although this model is based on an Formulated with Bentonite and Sodium
unweighted bentonite suspension, similar Carboxymethyl Cellulose,” J. Pet. Sci. Eng.
work on weighted muds may be carried out 57, 294-302.
in order to study the effect of added solids
on thixotropic behaviour. There is scope for 8. Cheng, D.C-H. and Evans, F. (1965),
further work to determine the temperature “Phenomenological Characterisation Of The
dependence of such behaviour, as well as its Rheological Behaviour of Inelastic
nature and effect in turbulent flows. Reversible Thixotropic and Anti-
Thixotropic Fluids,” Brit. J. Appl. Phys., 16,
ACKNOWLEDGMENTS 1599.
The author thanks M-I SWACO for 9. Moore, F. (1959), “The Rheology of
permitting the publication of this work. Ceramic Slips and Bodies,” Brit. Cream.
Soc. Trans., 58, 470.
REFERENCES
1. Jenekhe, S.A., Davis, H.T. and Scriven, 10. Zamora, M. and Lord, D.L. (1974),
L.E. (1989), “Cold Stage Scanning Electron “Practical Analysis of Drilling Mud Flow in
Microscopy of Bentonite Suspensions,” 38th Pipes and Annuli,” SPE 4976, 49th Annual
Annual Proceedings of Electron Microscopy Fall Meeting of the Soc. Pet. Eng.
Society of America, (Ed. G.W. Bailey), San 11. Herschel, W.H and Bulkley, R. (1926),
Francisco, California, 206. “Measurement of Consistency as Applied to
2. Dairanieh, I.S. and Lahalih, S.M. (1988), Rubber-Benzene Solutions” Proc. Amer.
“Novel Polymeric Drilling Mud Soc. Test. Mat., 26, 621.
Viscosifiers,” Euro. Polym. J. 24, 831.
3. Alderman, N.J., Ram Babu, D., Hughes,
T.L. and Maitland, G.C. (1988), “The
Rheological Properties of Water-Based
Drilling Fluids,” 10th International Congress
on Rheology, Sydney, August 14-19.