Professional Documents
Culture Documents
This paper was prepared for presentation at the SPE Western North American and Rocky Mountain Joint Regional Meeting held in Denver, Colorado, USA, 16–18 April 2014.
This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessar ily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohi bited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.
Abstract
In drilling fluid rheological characterization and hydraulics modeling, selecting a proper rheological model and obtaining
rheological parameters using viscometers are critically important. Bingham Plastic, Power Law, and Yield Power Law are the
most commonly used standard models for drilling fluid rheology since they are mathematically simple to implement.
However, due to the complex nature of drilling fluids, these models do not often fit well to the rheological data. This leads to
a significant error in drilling fluid hydraulics modeling. While searching for models based on the physical interaction
between the various components of the drilling fluid, it is very practical to find simple mathematical functions that best fit to
any experimental data obtained from field viscometers. This will improve the accuracy of hydraulics modeling, in particular
for time dependent drilling fluids.
In this research, cubic splines have been used to fit the experimental data obtained from field viscometers. A generalized
hydraulics model is presented to calculate pressure drop in pipe and concentric annulus with detailed sample calculations. A
numerical simulator has also been developed to assist the calculations. The results obtained from the model have been
validated with experimental data to verify the effectiveness of the proposed model.
The model introduced in this paper is the generalized form of all the other commonly utilized models and can be
implemented for any fluid. The main advantage of the model is that it can capture all the complex rheological response of the
fluid. Hence, all readings obtained from the viscometers can provide valuable input for the hydraulic model. Also, the
numerical solution is stable and straightforward to implement. The model provides drilling engineers a simple and powerful
method to accurately predict the pressure distribution along the wellbore. This is highly important in drilling fluid hydraulics
program optimization and well control operation, especially in deep-water and arctic environments.
Introduction
The relationship between stress and deformation (shear strain, shear rate) should be written in tensor form since these
quantities are tensor quantities. However, in drilling fluid applications, uniaxial shear, or simple shear, is commonly
encountered. The constitutive equation in a single direction is widely used. For simplicity, in the next section, the relation
between shear stress and shear rate, and the flow curve equation, is written in one-dimensional form.
In the early days, the linear correlation, the Newton model, was the most popular model. In simple shear, the response of
a Newtonian fluid is described by a linear relationship between the applied shear stress and the rate of shear. After
advancements of modern viscometers and rheometers, researchers showed that complex fluids do not obey this linear
Newtonian relationship. These fluids are classified as non-Newtonian fluids. Non-Newtonian fluids are those for which the
flow curves are not linear, i.e. the viscosity is a complex function of several factors including shear rate and shear history.
Non-Newtonian fluids are often subdivided in three groups. These are time independent, time dependent, and viscoelastic
fluids (Wilkinson, 1960). Various correlations have been developed to model the rheological response of non-Newtonian
fluids. The most common models are the Bingham Plastic, Power Law, Herschel-Bulkley, Carreau-Gahleitner, Casson-
Steiner, Casson, and Ellis models (Figure 1). A statistical review of these models for the petroleum industry applications has
been published by Weir et al. (1996). One of the common advantages of these models is that they have a simple mathematical
form. In other words, the number of model parameters is limited and can be obtained from a small number of experiments. It
is sometimes easier to obtain analytical solutions in hydraulics problems. However, these models do not always give a good
fit to the rheological data and this ill-fates the hydraulics calculations.
2 SPE 169527-MS
Due to the complex components of drilling fluids and the diversity of drilling fluid systems, drilling fluids are normally
classified as non-Newtonian fluids. Also, drilling fluids often show time dependent properties and behave viscoelastically
under certain conditions (Bui, 2012). This explains why rheological characterization of drilling fluids at different
measurement conditions and procedures gives different results. The time dependent properties are resulted from the
interaction among various components of the drilling fluid and the complex response of each component under different
shearing conditions. These components are often different in size, chemical activity, specific gravity, and surface energy. One
more important phenomenon that often makes the rheological response of the drilling fluid complex is gelation. To fulfill its
complex functions, such as cuttings transport and solid suspension, structure build up or gelation is a desired property of the
drilling fluids. The gel formation, when the fluid is at rest, helps to keep the solid particles from settling. Gel structure
formation helps to prevent fluid invasion into the formation and lost circulation problems (Bui, 2012). Therefore, the
response of drilling fluid in rheological measurement is very complex. Experimental data are repeatable with some degree of
certainty if experimental procedure and all of the conditions are the same. Also the flow curve of drilling fluids is very
complex, and simple rheological models do not often fit well to the experimental data.
In the petroleum literature, many attempts have been made to better characterize the rheological response of the drilling
fluids. Most of the rheological models that are currently used in Petroleum Engineering are inherited from Chemical and
Mechanical Engineering. There are only some models developed by the drilling industry for drilling fluid modeling including
the work of Robertson and Stiff (1976) and Pilehvari et al. (2001).
Robertson and Stiff (1976) presented an improved mathematical model for the drilling fluids and cement slurries. Their
model has the following form:
A( C ) B (1)
Pilehvari et al. (2011) presented a model in the Rational Polynomial form as,
P P P2 2 .... Pk k
o 1 (2)
1 Q1 Q2 2 .... Qm m
Beside models formulated by the drilling industry, other research focused on the applicability of available models for
modeling of the drilling fluids. Okafor and Evers (1992) provided a comparison of several models including Robertson-Stiff,
Power Law, Bingham Plastic with the experimental data. Bailey and Peden (2001) showed the applicability of Sisko model.
Wang et al. (2010) show that modified Sisko model, or four-parameter model, also fits well to the experimental data.
It is evident that most available rheological models are originally developed as empirical correlations. They are
mathematical equations that give a good fit to the experimental data. The models with more parameters often fit the
experimental data better than the models with fewer parameters. There is also limited effort to derive constitutive equations
based on the concentration and the physical and chemical interactions between the different components of the fluids. This
approach can be used for a simple mixture in the absence of the experimental equipment for measurements. However, it is
not very practical for drilling fluid applications as the drilling fluids have a very complex nature and their rheological
SPE 169527-MS 3
characteristics are strong functions of both pressure and temperature. Hence, making a best use of the available experimental
data is a more practical approach.
In this research, cubic splines have been used to fit the rheological data from viscometer measurements to construct the
flow curve. From the constructed flow curve, a hydraulics model is formulated using the Rabinowitsch-Mooney method. The
result is then validated with measured data to verify the applicability of the introduced model.
Model Formulation
Cubic Spline Determination
A cubic spline is a sufficiently smooth cubic function that is piecewise-defined, and possesses a high degree of smoothness at
the points where the polynomial pieces connect. The rheological data from a rotational viscometer can be extrapolated using
cubic splines as,
1 c01 c11 c21 2 c31 3 , o 1
c12 c12 c22 c32 , 1 2
2 3
2 (3)
...
c c c 2 c 3 , n1 n
n 0n 1n 2n 3n
The equation above contains 4n unknowns that can be determined from the viscometer measurements. The coefficients
of each equation can be obtained using the following constraints:
The splines have to pass through all points:
i i i1 i c0i c1i c2ii2 c3ii3 c0i1 c1i1 c2i1i2 c3i1i3 (4)
The slope or the first derivative of the two consecutive cubic functions at the common point must be equal,
i i i1 i
i' i i'1 i c1i 2c2ii 3c3ii2 c1i1 2c2i1i 3c3i1i2 (5)
The second order derivative of two functions at the common point must be equal,
2 i i 2 i1 i
i" i i"1 i 2c2i 6c3ii 2c2i1 6c3i1i (6)
2 2
For natural cubic splines, the second order derivative at the two points is zero,
2 1 0
2 2c21 6c31o 0
(7)
2
n n 2c 6c 0
2 2n 3n n
From Equations 4 thru 7, we can obtain 4n equations to solve for the 4n unknown coefficients. Hence, the cubic splines
can be constructed if rheological data are available.
In the rheological data collected using six-speed field viscometers, the shear rate range between two consecutive speeds
is not uniform. There is more data at the lower shear rates than that at the higher shear rates. Natural cubic splines often result
in overshooting in the low shear rate range. We often observe the overshooting at a shear rate range from 10.22 to 170.3 [s-1]
corresponding to rotational speed of 6 to 100 [RPM] on field viscometers. To avoid overshooting, more data should be
collected using measurement devices with more rotational speeds. One simple method in the absence of these devices is to
use constrained cubic splines. To construct constrained cubic splines, the approach introduced by Kruger (2012) is used in
this paper. We eliminate the requirement of equal second order derivatives and replace it with the specified first order
derivatives defined numerically as,
' 2 i i1 i1 i
i i
i1 i i i1 i i1 i1 i
' 3 1 0 1' 1
1 0 (8)
2 1 0 2
' 3 n n1 n' n
n n
2 n n1 2
where the second order derivatives are defined through the first order derivative as,
2 i i 1 6 i i 1 2 i' i 2 i' i 1
2 i i 2
"
2 i i
i i1 2
i i1
From a set of experimental data i , i , the first order derivative can be evaluated using Eq. 6. The second order
derivative is evaluated using Eq. 8. Then spline coefficients can be determined using Eq. 7.
Q v(r )drd
0 0
(12)
Assuming no wall slippage, and changing variables, we can arrive at the Rabinowitsch-Mooney equation correlating wall
shear stress and flow rate for pipe flow of any fluids as shown on Eq. 13.
w
R 3 2
w2 0
Q d (13)
Substituting the equation for shear stress (Eq. 3) into Eq. 13 and using shear rate as the primary variable, flow rate is
calculated from shear rate as,
0
c01 c11 c21 c31 d c01 c11 c21 c31
2 3 2 2 3
0
2
c02 c12 c22 2 c32 3 d c02 c12 c22 2 c32 3
2
R 3 (14)
Q 1
c0i c1iw c2iw2 c3iw3 2
...
w
c0 i c1i c2 i c3i d c0 i c1i c2 i c3i
2 3 2 2 3
i
0
c01 c11 c21 c31 c11 2c21 3c31 d
2 3 2 2 3
0
2
c02 c12 c22 2 c32 3 c12 2c22 2 3c32 3 d
2
R 3 (15)
Q 1
c0i c1iw c2iw2 c3iw3 2
...
w
c0i c1i c2i c3i c1i 2c2 i 3c3i d
2 3 2 2 3
i
SPE 169527-MS 5
Since all coefficients in the equation above are determined from the rheological data, the integrant in the right hand side
is a polynomial. For an assumed value of the wall shear stress, the integral can be evaluated analytically. By varying the
value of wall shear rate until two sides are equal, this equation can be solved to obtain the shear rate at the shear wall.
Substituting shear rate into the equation for shear stress (Equation 11), wall shear stress can be obtained. The frictional
pressure gradient is calculated from the correlation between the frictional pressure gradient and the wall shear stress.
Annular Flow
For a concentric annulus, using narrow slot approximation approach, a concentric annulus can be represented by a rectangular
slot with the height, H, and the width, W, defined as,
Do Di
H and W D o
Di (16)
2 2
The narrow slot approach only accounts for the geometrical correspondence between annular and rectangular slot, but
the variation of flow properties on the wall such as wall shear rate and wall shear stress is not considered. For a real annular
flow, the wall shear stress and wall shear rate on two walls, outer and inner walls, are not the same. However, in the narrow
slot approximation, the wall shear rate and wall shear stress on two rectangular walls are the same. For a very narrow slot, the
wall shear stress and wall shear rate on both inner and outer walls are relatively equal. In this case, the narrow slot
approximation is very effective. For an annulus with high diameter ratio (D o/Di), this approach leads to significant errors. To
consider both geometrical constraints and wall properties, we introduce a shape correction factor (b) and define the
corresponding rectangular slot height and width as following:
Do
bDi
Do
H b and W bDi (17)
2 2 b
b is larger for higher diameter ratio (Do/Di) and approximately equal to unity for a very narrow slot.
By measuring the pressure gradient in annular flow, average wall shear stress w can be calculated. Also from the
measurement of wall shear stress at the outer and inner walls wo and wi , the shape correction factor can be determined
from the force balance equation (Eq. 18). This will be presented in another publication.
Do D
bDi L w Do wo Di wi L or o bDi w Do wo Di wi (18)
b b
Similar to pipe flow, using force balance, we obtain the following correlation between the wall shear stress and the
pressure gradient:
P H P 2
w or w (19)
L 2 L H
Flow rate can be obtained from:
W H W H /2
With the assumption of no wall slippage, the Rabinowitsch-Mooney equation for the rectangular slot has the following
form:
w
WH 2
Q
2 w2 0
d (21)
Substituting the equation for shear stress into Equation 21 and using shear rate as the primary variable, flow rate is
obtained in Equations 22 and 23.
0
c01 c11 c21 c31 d c01 c11 c21 c31
2 3 2 3
2 0
Do D 2
bDi o bDi c02 c12 c22 2 c32 3 d c02 c12 c22 2 c32 3
1
b b (22)
Q
16 c0 i c1i w c2 i w2 c3i w3 2
...
w
c0 i c1i c2i 2 c3i 3 d c0 i c1i c2i 2 c3i 3
i
6 SPE 169527-MS
0
c01 c11 c21 c31 c11 2c21 3c31 d
2 3 2 3
0
2
2
Do D
bDi o bDi c02 c12 c22 2 c32 3 c12 2c22 2 3c32 3 d
1
b b (23)
Q
16 c0 i c1i w c2i w2 c3i w3
2
...
w
c c c 2 c 3 c 2c 2 3c 3 d
0 i 1i 2 i 3 i 1i 2 i 3 i
i
Again, all coefficients in the equation above are known, the wall shear rate can be evaluated numerically, and wall shear
stress and frictional pressure gradient by solving Eq. 23.
The generalized Reynolds Number is calculated through the wall shear stress from Eq. 24.
8v 2
Re (24)
w
Computational Procedure
By assuming an initial value of the wall shear rate, typically a small value, the integral on the right hand side can be
evaluated analytically. Then, the right hand side is compared with the given flow rate. If two values are equal within a
defined tolerance, then the calculation is ceased. Else, the wall shear rate is increased and the right hand side is evaluated
again. This procedure is repeated until convergence is achieved (Fig. 1). Then the wall shear stress, pressure gradient and
generalized Reynolds Number can be computed from the wall shear rate using Eq. 24.
Rheological Data
Start with 𝛾𝑤 = 0
|Qcal-Qgiven|≤Tolerace No 𝛾𝑤 = 𝛾𝑤 + ∆𝛾𝑤
Yes
END
Sample Calculation
In this section, we present a detailed example to demonstrate the calculation procedure of the model introduced. This
example includes both pipe and annular flows. The pipe diameter used is D = 0.1143 [m] and the annulus with inside
diameter is Di = 0.127 [m] with an outside diameter of Do = 0.254 [m]. The drilling fluid density is 1200 [kg/m3]. The
rheological data from lab measurement is presented in Table 1.
SPE 169527-MS 7
From the rheological data provided in Table 1, the following natural cubic splines can be constructed to model the flow
curve of the drilling fluid.
0 3.066 0.196046 0.001172 2 7.6 *10 5 3 , 0 5.11
1 3.066 0.196046 0.001172 7.6 *10 , 5.11 10.22
2 5 3
From Eq. 25, the flow curve can be drawn. A comparison of the natural cubic spline model and other models is presented
in Figure 3. It is evident that the coefficients corresponding to and are relatively small. These coefficients become
2 3
very small at very high shear rate indicating that the non-linear part is small and viscosity is nearly constant at high shear
rates. This agrees with the observation that Newtonian behavior is often observed at very low and very high shear rate. In
other words, the cubic spline model can predict very well the Newtonian response at very high shear rates.
50
Experimental Data
45 Natural Cubic Spline
Constrain Cubic Spline
40 Binham Plastic
Yield-PowerLaw
35 Power Law
30
Shear Stress [Pa]
25
20
15
10
0
0 200 400 600 800 1000
Shear Rate [1/sec]
After the flow curve is formulated, the aforementioned computational procedure is used to calculate the wall shear rate
for a given value of the flow rate. Then, wall shear stress and pressure gradient are evaluated. The frictional pressure gradient
for different flow rates with corresponding Reynolds Number is shown in Figure 3 for pipe flow and in Figure 5 for annular
flow.
One more important observation is the overshooting of the natural cubic spline curve on the shear rate ranges from 10.22
to 170.3 [s-1]. This results in overshooting on the wall shear stress versus the wall shear rate curve (Figures 5 and 6), since
the wall shear stress versus the wall shear rate curve is the same as the flow curve on Figure 3. In this case, using constrained
cubic spline will resolve problem.
For annular flow, with the diameter ratio, Do/Di = 2, the shape correction factor b = 1.085 is used.
8 SPE 169527-MS
Figure 5 –Wall shear stress as a function of wall shear rate for pipe flow.
Figure 7 - Wall Shear Stress as a Function of Wall Shear Rate for Annular Flow.
Model Validation
The experimental data used in this section if from Okafor and Evers (1992). The pipe diameter is D = 2 [in] and the annulus
has inside diameter of Di = 1.5 [in] with an outside diameter of Do = 3 [in]. Two drilling fluid experiments (fluid type A and
fluid type B) have been used. The density of fluid type A is 1066.2 [kg/m 3] and that of fluid type B is 1036.3 [kg/m3]. The
rheological data are presented in Figure 8 for fluid type A and Figure 11 for fluid type B. For annular flow, the shape
correction factor b = 1.085 is used for the diameter ratio D o/Di = 2.
For Fluid Type A
25
Experimental Data
Cubic Spline
20
15
Shear Stress [Pa]
5.1 0.958
10.2 1.437
5 170.3 6.226
340.7 10.537
511 13.89
1022 22.99
0
0 200 400 600 800 1000
Shear Rate [1/sec]
Figure 8 - Flow curve from FANN 32 viscometer data (fluid type A) data is after Okafor and Evers (1992).
10 SPE 169527-MS
From the rheological data, the constrained cubic splines are constructed in Equation. 26.
0 0.479 0.093922 5.6 *1017 2 6.49 *1018 3 , 0 5.11
1 00479 0.045374 0.014279 9.3 *10 , 5.11 10.22
4 3
2
0.955176 0.049154 1.9 *104 2 5.06 *107 3 , 5.11 10.22
2
3 1.552198 0.026436 1.2 *10 3.6 *10 ,
(26)
5 2
8 3
10.22 170.3
1.6389 0.055469 7.5 *105 2 5.06 *108 3 , 170.3 510.9
4
5 3.420947 0.02272 5.2 *106 2 1.71 *109 3 , 510.9 1021.8
6 4.7961 0.0178 3.63 *10 5.2 *10 , 1021.8
9 2 13 3
Since the rheological data fit quite well to the Yield Power Law and Sisko models, these models can also provide a
relatively good prediction as compared to the cubic spline approximation. The pressure gradient for the pipe flow predicted
by Bailey and Peden (2001) using Sisko model is also presented here for a comparison (Figure 9).
TABLE 2 - COMPARISON OF OUR PREDICTION WITH MEASURED DATA AND BAILEY AND PEDEN PREDICTION
FOR PIPE FLOW (FLUID TYPE A)
Cubic Spline Approximation Measured Bailey and Peden
Flow
Pressure
Velocity Wall Shear Wall Shear Predicted Pressure Generalized Predicted Pressure Generalized
Gradient
[m/sec] Rate [1/sec] Stress [Pa] Gradient [Pa/m] Reynolds Number Gradient [Pa/m] Reynolds Number
[Pa/m]
0.202 39.8 2.637638 207.6881 148.511 189.7384 206.3246 130
0.409 77.4 3.837873 302.1947 418.4342 283.4229 305.9327 361
0.562 105.5 4.58644 361.137 661.1015 341.2011 368.3587 568
0.724 134.4 5.302695 417.535 948.9684 405.2675 428.6886 809
0.914 165.6 6.098549 480.2007 1315.034 469.9717 494.3953 1119
1.076 190.6 6.779597 533.8265 1639.426 522.7376 547.6169 1400
1.382 239 8.067948 635.2715 2272.606 644.0354 640.9368 1971
1.448 249.6 8.342397 656.8816 2412.779 669.99 699.9909 2101
It can be observed that the cubic spline model provides a very good fit to experimental data. Bailey and Peden (2000)
predicted the flow at velocity of 1.448 [m/sec] that is still in the laminar regime. The cubic spline model introduced here
shows that it is already out of the laminar regime, with Re =2413. This explains why measured data is deviated from the
value predicted by Bailey and Peden (2000) (Figure 8). Even though in their original publication Okafor and Evers (2001)
considered this point to be laminar, the results presented here verify that it should be in the transient regime.
The constrained cubic spline rheological model for this experimental data is presented in Eq. 27. The predicted pressure
gradient and a comparison with the measured data for both pipe and annular flows are shown in Table 4.
0 8.61845 0.23471 4.4 *10 16 2 5.6 *10 17 3 , 0 5.11
1 8.61845 0.04829 0.05483 3.58 *10 , 5.11 10.22
2 3 3
Figure 9 – frictional pressure gradient for pipe flow (fluid type A).
Figure 10 – Frictional pressure gradient for annular flow (fluid type A).
30
Experimental Data
Cubic Spline
25
20
Shear Stress [Pa]
15
10
0
0 200 400 600 800 1000
Shear Rate [1/sec]
Figure 11 - Flow curve from FANN 32 viscometer data (fluid type B)
12 SPE 169527-MS
TABLE 4 - COMPARISON OF THE MEASURED DATA FOR PIPE AND ANNULAR FLOW (FLUID TYPE B)
Pipe Flow Annular Flow
Figure 12 – Frictional pressure gradient for pipe flow (fluid type B).
Acknowledgments
The authors acknowledge the support from the sponsors of the Unconventional Natural Gas and Oil Institute (UNGI)
CIMMM Consortium and Petroleum Engineering department at Colorado School of Mines.
Nomenclature
A, B, C = Robertson-Stiff model parameters. Greek Symbols
b = Shape correction factor. = Shear rate, 1/sec.
ci = Spline coefficient. w = Wall shear rate, 1/sec.
Di = Inside diameter, m.
Do = Outside diameter, m. w = Average wall shear rate, 1/sec.
dp/dL = Frictional pressure gradient, Pa/m. = Incremental shear rate, 1/sec.
H = Rectangular slot height, m. = Shear stress, Pa.
R = Pipe radius, m. w = Wall shear stress, Pa.
r = Radial position, m.
Re = Generalized Reynolds number. wi , wo = Wall shear stress on the inner and outer wall
Pi, Qi = Rational polynomial model parmeters. respectively, Pa.
W = Rectangular slot width, m. w = Average wall shear stress, Pa.
References
1. Bailey, W.J. and Peden, J.M, 2000, A Generalized and Consistent Pressure Drop and Flow Regime Transition Model for
Drilling Hydraulics, paper SPE-62167-PA, Journal SPE Drilling and Completion 15(1), pp.44-56
2. Bui, B., 2012. Determination of Viscoelastic Properties of Drilling Fluids. MSc. Thesis, the University of Tulsa, Tulsa,
Oklahoma, USA.
3. Kruger C.J.C, 2002, Constrained Cubic Spline Interpolation for Chemical Engineering Applications.
4. Okafor, M. N. and Evers, J. F., 1992, Experimental Comparison of Rheology Models for Drilling Fluids, Paper SPE
24086 presented at the Western Regional Meeting held in Bakersfield, California, USA.
5. Pilehvari, A., Serth, R., and Lagad, V., 2001, General Hydraulics Calculation Method Using Rational Polynomial Model.
Paper SPE 71403 presented at the 2001 SPE Annual Technical Conference and Exhibition held in New Orleans,
Louisiana, USA.
6. Wang, G. and Fan, H.H., 2010, Utility Calculation Method of New Four-Parameter Rheological Model for Drilling
Hydraulics, Paper SPE 130421 presented at the SPS/SPE International Oil and Gas Conference and Exhibition held in
Beijing China.
7. Weir, I.S. and Bailey, W.J., 1996, A Statistical Study of Rheological Models for Drilling Fluids, paper SPE-36359-PA,
SPE Journal 1(4), pp.473-486.
8. Wilkinson W.L., Non-Newtonian Fluids, Peramon Press, 1960.
9. Robertson, R.E., Stiff Jr., H.A., 1976, An Improved Mathematical Model for Relating Shear Stress to Shear Rate in
Drilling Fluids and Cement Slurries, Paper 5333-PA, SPE Journal, pp. 31-36.