Académique Documents
Professionnel Documents
Culture Documents
Stage:
Page: 279
Total Pages: 16
Summary
A dynamic stiff-string torque-and-drag (T&D) model is presented
that assumes steady-state motion of the drillstring as its basis for
calculations. Results are compared with previously published
T&D models that are based on static equilibrium. The novelty of
the new dynamic model is the ability to solve T&D operations of
the entire drillstring from bit to topdrive in reasonable time by use
of standard engineering computers.
The new approach is modeled after a 3D dynamic model of
drillstring and bottomhole assembly in an elastic borehole. It considers bending stiffness, torsional stiffness, contact forces, and
friction with localization of contact points. A numerical method is
described that has proved to have excellent convergence. Complete governing equations are provided, and the method is
described in detail to permit readers to replicate results.
The dynamic model is compared with two static stiff-string
models. Comparisons are also provided for three conventional
soft-string models, including the Lubinski-Paslay-Cernocky bending-stress-magnification factor. Four field case studies are presented for horizontal wells. One well is short radius with dogleg
severity greater than 50 /100 ft, and three wells are unconventional shale wells with doglegs up to 15 /100 ft. Predictions for
surface T&D up/down for the new dynamic stiff-string model are
compared with the static stiff- and soft-string models. In many situations modeled, the top-level results for surface T&D up/down
are close enough for all six models to be within the uncertainty
range associated with the commonly used, lumped-parameter friction factor. However, some major differences in hookload for sliding and slackoff operations are observed, which are shown to be
caused by differences in location and magnitude of contact force
between the drillstring and wellbore. Further, significantly lower
surface torque is predicted by the new dynamic stiff-string model
compared with other models for one case history because of lower
contact forces in the vertical section of the well. In fact, the key
finding of this paper is that major differences are observed for
contact forces for the new dynamic stiff-string model compared
with all five other models, including the two static stiff-string
models. These differences in contact forces are most significant
when the drillstring has helically buckled or when doglegs in the
wellbore are high. Contact forces have a large impact on local
stress behavior, which is important for predictions of casing
and drillpipe wear, drillstring fatigue, and failure points in the
drillstring.
Although several previous papers have published stiff-string
models, there is no industry-standard formulation. The main problem holding back the development of an industry-standard stiffstring model is perhaps the complexity of the numerical algorithm
and substantial running time. To address this problem, some previous stiff-string models account for bending stiffness of the drillstring but not for radial clearance, whereas others appear to model
only portions of the drillstring as stiff. The new stiff-string model
accounts for bending stiffness and radial clearance for the entire
drillstring while still giving reasonable computational times. For
This paper (SPE 163566) was accepted for presentation at the 2014 IADC/SPE Drilling
Conference and Exhibition, Fort Worth, Texas, USA, 46 March 2014, and revised for
publication. Original manuscript received for review 29 April 2014. Paper peer approved 4
August 2014.
Shortcomings of Soft-String Models. The fact that it is impossible to permit radial clearance is a critical shortcoming because
most of the drillstring comprises jointed drillpipe (or heavyweight
279
Stage:
Page: 280
Total Pages: 16
drillstring is either not moving or moving without any time dependency. Static means the drillstring is not moving. Somewhat
in conflict with this view, the classic soft-string models of Johancsik et al. (1984) and Sheppard et al. (1987) were classified by
Mason and Chen (2007) as steady state, meaning that transient
effects were ignored and the drillstring was assumed to be moving
in a steady manner rather than being static. Regardless of whether
the soft-string model permits any motion of the drillstring, the
fact is that velocity terms do not appear in the formulation.
Indeed, one of the industry-available programs compared in this
paper does not permit the user to enter any velocity values (such
as axial speed or rotary speed) for both its soft-string and stiffstring models. The other industry-available program does not permit user input for rate of penetration, or axial speed, for sliding
operations for both its soft-string and stiff-string models. So, it is
clear that these programs do not consider essential elements of a
dynamic program. The only manifestation of axial drillstring
movement in the soft-string models is the addition of the drag
term, which is assumed to be the product of the coefficient of friction multiplied by the contact force, again with an unspecified velocity. This approach does not really entail a dynamic drillstring
model but rather is a useful way to include friction. Moreover, the
torque term, which is frictional resistance to rotary motion, is the
same frictional drag term multiplied by a somewhat unknown (or
at least arbitrary) moment arm. Consequently, we refer to the
industry-available programs as static.
Stiff-String Models. To overcome the shortcomings of softstring models, new and more complex models were developed.
Generally these other models are lumped into a fairly broad category known as the stiff-string model.
Inclusion of Bending Stiffness. Ho (1988) discussed the
shortcomings of the soft-string model and developed an improved
model that combines abottom hole assemblyanalysis in the
stiff collar section, coupled with an improved soft string model
for the remainder of the drillstring. The improvement offered is
that the bending stiffness of the drillstring is included but the drillstring is still assumed to be in continuous contact with the wellbore (Ho 1988).
Mitchell and Samuel (2007) developed an analytical model
similar to those of Ho (1986, 1988) and studied the classic softstring model. Mitchell (2008) offered improvements to the classic
soft-string model by replacing the commonly used minimum-curvature-interpolation method for directional surveys with a spline
model. Run times are stated to be extremely fast, less than 0.10
second on a modern personal computer. There is growing acceptance in the industry that spline interpolation provides a better
description of the likely drillstring configuration than minimum
curvature. However, Mitchells spline model assumed that the
drillstring matched the well-path curvature (using the improved
spline well path), thereby remaining in continuous contact with
the wellbore. Consequently, this model does not permit radial
clearance, which is an essential element of a holistic stiff-string
model.
Inclusion of Radial Displacement. Cernocky and Scholibo
(1995) stated that their development of an analytical model that
accounted for bending stiffness and radial clearance was to
address column buckling of casing. Details about the analytical
model are not provided, but extensive details on associated numerical modeling by use of Abaqus finite-element analysis (FEA)
are given in a separate reference (Cernocky and Scholibo 1994).
Their objective was to study casing damage caused by reservoir
compaction. The primary modeling focus was to simulate depletion of the reservoir pressure and corresponding loads and deformation of the casing. The local model of the well was used to
predict the reservoir loading on the casing, by use of nodal loads
instead of distributed loads on the casing. Central-processing-unit
(CPU) run times were extremely slow, about one CPU week on
an HP730 workstation (Cernocky and Scholibo 1994).
Rezmer-Cooper et al. (1999) used a finite-element stiff-string
model for both bending stiffness and radial clearance. Results for
September 2014 SPE Drilling & Completion
Page: 281
Total Pages: 16
X (North)
Z
Stage:
O
e3
e1
O
e2
Q
M(0)
Fig. 1Coordinate systems.
one case history showed the contact forces for stiff string were
significantly lower than for soft string (Rezmer-Cooper et al.
1999). The paper also states that their stiff-string model calculated
a much more accurate torque loss for the drillstring compared
with the soft-string model, but it did not say which was higher.
Menand et al. (2006) developed a stiff-string model of a complete drillstring that uses a numerical method that was stated to be
much faster than finite element. The drillstring is assumed to be a
beam element, and the borehole is assumed to be rigid and circular. The contact algorithm is begun by assuming no contact anywhere in the drillstring and then uses an iterative process to
introduce contact, one point after another.
Aslaksen et al. (2006) described FEA analysis of the drillstring
from bit to surface. The analysis ran on a Unix platform and provided time-based simulations of the drilling process that included
a laboratory-derived rock/bit-interaction model. The method modeled torsional, axial, and lateral vibrations and accelerations. The
paper points out solutions for the entire drilling system are
exceptionally time consuming and complex both in setup and
computational time. Although FEA modeling had been used as
early as 1943, the paper stated that the long computational time
had meant that relatively short lengths of drillstring had been analyzed. The general approach is described for this method, but sufficient details on the solution method are not provided to allow
readers to replicate the papers results.
McSpadden et al. (2011) carries out full stiff-string analysis of
casing by use of a purpose-built FEA code to calculate radial displacements of the casing centerline from the wellbore centerline.
The radial deflection of the casing string at each node allows for
determination of the true casing curvature or effective dogleg severity. The model was run in static mode although the same stiffstring-analysis model has also been used in dynamic mode to study
stresses on the coiled-tubing stack from wellhead to injector assembly (Smalley and Newman 2005). Application of the dynamic
model has also been made to analyze drillstring loads during jarring (Newman and Proctor 2009).
Summary. Two general types of mathematical models are in use
for T&D software: soft string and stiff string. The soft-string
model is widely available and represents the industry standard. It
is used because of the simplicity of its algorithm, rapid calculation
time on even the most basic laptop computer, and the sufficiently
accurate results obtained for many common drilling situations.
Stiff-string models have been developed to provide more-accurate analysis than soft-string models. It is generally believed that
00
jMr jj0 f n ;
00
1
M0 l; . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
qIp w
m
x T 0 fx ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
where r y jz; y, z projections of the drillstring axial line on
axes oe2 and oe3; m* effective mass, which includes mass of the
drillstring m and the added fluid mass inside the drillstring and in
281
the annulus per unit length; EI, GIp, AE bending, torsional, and
axial stiffness of the drillstring, respectively; qIp moment of
inertia of the drillstring; T AEx0 0:5jr0 j2 axial force; M
GIp w0 j1 torque; j a0 jh0 cosa a projection of the
well curvature onto plane oe2oe3; j1 cosah0 a projection of
the well curvature onto axis oe1; a, h the inclination angle and
the azimuth angle of the wellbore, respectively; w torsion angle;
fn fy jfz; fx, fy, fz projections of the external force per unit
length onto axes oe1, oe2, and oe3, respectively; l frictional torque per unit length; and j (1)1/2 imaginary unit.
Note the prime means a derivative with respect to arc length s
(Fig. 1) and the dot is derivative with respect to time t.
Major components of external forces and torque include
weight, contact force, and friction force:
f n wsina pr=jrj1 jkh vh =v; . . . . . . . . . . . . . . . 4
_
. . . . . . . . . . . . . . . . . . . . . . . 5
fx wcosa kx px=v;
l 0:5Dpkh ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
where w weight of the drillstring in mud per unit length;
_ circumferential
p contact force per unit length; vh 0:5Dw
velocity of a point on the drillpipe surface; v x_ 2 v2h 1=2 resultant velocity of a point on the drillpipe surface; D outer
diameter of the drillstring; and kh, kx friction factors in the circumferential and axial direction, respectively.
The model assumptions are
The contact force is dependent on a penetration of drillstring
into the wellbore wall: p p(d), where d (y2 z2)1/2 d0;
d0 0.5(Dh D) clearance between the wellbore wall and
the drillstring; and Dh hole diameter.
Dependence p(d) is quadratic-elastic (Tikhonov et al. 2006).
The circumferential velocity of a point on the drillpipe surface relative to its own axis is significantly higher than the
circumferential velocity of a point on the drillpipe surface
relative to the wellbore axis. This assumption permits the
simulation of drillstring-whirl cases.
Dahls elasto-plastic model is used as the friction model
(Tikhonov and Safronov 2011).
The drillstring and wellbore parameters can vary along the
measured depth. This feature permits modeling different drillstrings
and bottomhole assemblies in wells with varying wellbore diameter.
For field case histories presented here, we assumed gauge hole.
Different rig operations are determined by boundary conditions for Eqs. 1 through 3.
For Eq. 1, the inputs required are lateral displacements at
the top and bottom of the drillstring in the cross-sectional
plane of the well r(0) and r(L), and the corresponding first
and second derivatives, where L drillstring length.
For Eq. 2, the inputs required are torque on bit,
Q GIp w0 O, and the drillstring rev/min, angular velocity
_
X wL.
For Eq. 3, the inputs required are weight on bit, P T(0),
_
and rate of penetration or axial speed, vr xL.
The equations of motion, Eqs. 1 through 3, are numerically
integrated by use of the method of lines (Tikhonov et al. 2006).
Eqs. 1 through 3, which are of the second order with respect to
time, are converted to six equations of the first order (i.e., in
Cauchys form). Partial derivatives with respect to length are
approximated by nonuniform finite differences on mesh along the
arc length, s. The resulting system of ordinary-differential equations is integrated by use of the Runge-Kutta explicit method.
Initial conditions for integration of the system of Eqs. 1
through 3 can be arbitrary, although reasonable assumptions are
used to improve convergence. For example, one reasonable initial
condition assumes that only all tool joints are in contact with the
wellbore wall. The initial conditions for the drillstring in the wellbore are specified in the first stage of the integration process. This
stage is used for identification of the natural position of the drillstring inside the hole, which includes locating all contact points in
a 3D well under gravitational force, friction, and contact forces.
282
Stage:
Page: 282
Total Pages: 16
00
00
00
8
M0 0:5Dpkh ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
T 0 wcosa kx pvr v: . . . . . . . . . . . . . . . . . . . . . . . 10
The steady-state condition was considered to have been
achieved when the kinetic energy at any point along the drillstring
(and neglecting rotation at the input rev/min) was less than 0.01 J.
The conventional soft-string equations of T&D can be obtained from Eqs. 9 and 10. Eq. 8 represents the equilibrium of
forces acting normal to the borehole wall and determines the distribution of the contact force along the length of the drillstring,
p(s). It is worth noting that the obtained solution considers the
drillstring-bending stiffness and allows the user to determine axial
force and torque, as well as bending moment, Mb (Eq. 11), and
contact force, R (Eq. 12):
00
Mb 0:5Djr j; . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
September 2014 SPE Drilling & Completion
Stage:
Page: 283
Total Pages: 16
Contact Force per Joint vs. per Node. Soft-string models mathematically assume continuous contact of drillstring with casing/
wellbore. This assumption prevents soft-string models from predicting the actual location of contact along the drillstring. Furthermore, to calculate contact forces, an additional assumption is
usually made that only tool joints are in contact. In other words,
the contact force is conventionally lumped per 31-ft joint for
reporting purposes.
The new stiff-string model makes no assumptions regarding
location of contact. Instead, it calculates the force on each node,
which for a beam model can be as close as 3 ft. As the following
case histories show, the new stiff-string model predicts that in
some cases additional contact occurs on the pipe body between
tool joints, whereas in other cases some tool joints are not in contact at all. The capability to predict local contact behavior is important for analysis of casing and drillpipe wear and drillstring
fatigue. It also helps identify failure points, thus improving drillstring operational reliability.
T&D Analysis Results: Comparison and Discussion. The comparison of hookload and surface torque calculated by different
models for basic rig operations is shown in Table 3. At first
glance, the comparison of T&D results indicates good agreement
of results among programs. But a closer examination of the
operations of rotating off bottom (ROB) and sliding reveals some
important distinctions among programs. Furthermore, it is noteworthy that only the new stiff-string and new soft-string models
are capable of modeling reaming operations, as shown in Table 1.
ROB. The axial-force distribution along the drillstring while
ROB is at TD shows good agreement for the results obtained by
all models (Fig. 3a). This figure also shows that surface results for
R pDL: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
0
3,500
3,000
100
3,000
1,500
1,000
500
0
4,000
500
200
200
400
600
800
1000
N+/S (ft)
400
2,000
TVD (ft)
TVD (ft)
2,000
N+/S (ft)
2,500
400
1,5001,4001,2001,000 800 600 400 200
E+/W (ft)
200
5,000
1,200
6,000
0
1,400
0
4,000
2,000
1,500
1,000
500
0
5,000
500
1,000
800
6,000
700
600
200
100
7,000
E+/W (ft)
(a)
2,500
3,000
200
350
1,000
3,000
150
300
800
3,500
2,000
250
600
4,000
1,000
50
200
TVD (ft)
1,000
4,000
N+/S (ft)
(b)
8,000
200
400
9,000
0
(c)
Fig. 2Well profiles: (a) Case 1, backward-nudge well; (b) Case 2, short-radius well; (c) Case 3, backward-nudge well.
September 2014 SPE Drilling & Completion
ID: jaganm Time: 20:25 I Path: S:/3B2/DC##/Vol00000/140027/APPFile/SA-DC##140027
283
DLS
DLS
Page: 284
Total Pages: 16
20
20
10
10
0
500
1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 5,000 5,500 6,000 6,500 7,000 7,500 8,000 8,500 9,000 9,500 10,000
120
7,000
100
6,000
80
5,000
Torque (lbfft)
Stage:
60
40
New Soft String
New Stiff String
Program A Soft String
Program A Stiff String
Program B Soft String
Program B Stiff String
20
0
20
0
(a)
1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 5,000 5,500 6,000 6,500 7,000 7,500 8,000 8,500 9,000 9,500 1,0000
4,000
3,000
2,000
1,000
1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000 10,000
MD (ft)
500
0
0
1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000 10,000
(b)
MD (ft)
Fig. 3Comparison of the T&D results for ROB operation, Field Case 1: (a) axial load and (b) torque distribution. DLS 5 dogleg severity in units of degrees per 100 ft.
20
DLS
DLS
20
10
0
5,000
1,500
10
0
500
1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 5,000 5,500 6,000 6,500 7,000 7,500 8,000 8,500 9,000 9,500 10,000
Contact Force (lbf)
1,500
New Soft String
New Stiff String/Joint
New Stiff String/Node
Program A Soft String
Program A Stiff String
Program B Soft String
Program B Stiff String
0
5,000
DLS
500
2,000
3,000
4,000
5,000 6,000
MD (ft)
7,000
8,000
9,000
10,000
500
1,000
2,000
2,500
3,000
3,500
1,500
2,000
MD (ft)
2,500
3,000
3,500
1,500
500
(c)
1,000
MD (ft)
0.4
0
0
6,500
6,000
5,500
0.2
1,000
6,500
500
(b)
1,000
1,000
6,000
5,500
New Soft String
New Striff String/Joint
New Striff String/Node
Program A Soft String
Program A Striff String
Program B Soft String
Program B Striff String
500
1,000
(a)
Fig. 4Contact force during ROB operation for Field Case 1 along: (a) entire drillstring, (b) build section, and (c) top 2,500 ft of vertical section.
hookload are in good agreement for all models (within 4%). The
surface torque (Fig. 3b) for ROB has somewhat larger differences
among programs, with a maximum difference of approximately
1,000 lbf-ft for surface torque (15%). One of the main reasons for
the observed difference is the fact that contact-force distributions
for different models are not identical, as shown in Fig. 4. The biggest difference in contact force occurs for the Program A stiff
string, which predicts much higher values in the build, as shown
in Fig. 4b.
Stage:
Page: 285
Total Pages: 16
Table 3Field Case 1: T&D results comparison for hookload and surface torque.
Second, it can also be seen that the new stiff string predicts no
contact in the build for more than two whole joints at 5,500 ft, as
shown by the red line in the close-up of the build in Fig. 4b. For
the rest of the well, the new stiff string predicts that contact occurs
only on the tool joints. This is shown by the agreement between
the orange bars (which represent contact-force-per-node results)
and the red line (which represents contact force per joint). None
of the other models were able to capture this behavior.
Sliding. The sliding operation at TD in this well shows several noticeable differences between the programs (Fig. 5). For the
axial-load distribution (Fig. 5a), Program As stiff string stands
out as the lowest hookload. This is caused by the larger amount of
helical buckling predicted by the Program A stiff string, which
can be observed from the higher contact forces for the light blue
curve in the build section in Fig. 5b, and particularly in the closeup in Fig. 5c. When helical buckling occurs, the program dramatically increases contact forces. All programs, except both models
of Program B, predict the onset of the drillstring helical buckling
just above the build section. Program B predicts sinusoidal buckling but not helical buckling. This can be seen by the lower contact forces for the green curves (for Program B soft and stiff) in
Fig. 5b and particularly in the close-up in Fig. 5c.
Comparing contact forces predicted by the new stiff-string
model and the other programs, we note that distribution of contact
forces (lumped per joint) are in a fairly good agreement with Program B and the new soft-string model, with the only exception
being the prebuild interval. Although lumped contact forces in the
high-dogleg section (red curve in Figs. 5b and 5c) are close to the
other predictions, the contact forces per node (orange bars in Figs.
5b. and 5c) in the build and high-dogleg sections are much lower
than the lumped-per-joint contact force, because of multiple
instances of drillpipe/body contact. The new stiff-string model
gives greater insight on actual drillstring-contact-point prediction
by being able to predict helically buckled drillstring position in
the hole. In Fig. 5d, the position of the drillstring predicted by the
new stiff string is shown to be helically buckled in the wellbore
just above the build section. The upper part of the figure shows a
side view of the drillstring, with some tool joints contacting the
top of the wellbore whereas others contact the low side. The lower
part of the figure provides a top view of the drillstring with some
tool joints contacting either side of the wellbore. The color of the
tool joint represents the intensity of the contact force, as shown by
the code bar at the bottom of the figure, with red being the highest
values. Let us recall that buckling in the three soft-string models
is calculated by use of a separate equation that is bolted on to
the T&D analysis after its completion, as described earlier in the
Introduction. However, this is not the case with the new stiffstring model, which calculates the location of each contact as well
as the contact force at each location.
Case 2: Short-Radius Well. Case 2 is a short-radius horizontal
well drilled in Texas with a maximum dogleg close to 50 /100 ft,
as shown in the upper part of Figs. 6a, 6b, and 6c. The build starts
at 1,300 ft and ends at 1,500 ft, where the well becomes horizontal. The well remains horizontal until reaching total depth at 2,830
ft, with doglegs between 0.5 and 6 /100 ft (Fig. 6c). The entire
20
DLS
DLS
20
10
0
0
500
1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 5,000 5,500 6,000 6,500 7,000 7,500 8,000 8,500
5,000
60
20
0
20
4,500
4,000
Contact Force (lbf)
40
10
0
1,000 1,500 2,000 2,500 3,000 3,500 4,000 4,500 5,000 5,500 6,000 6,500 7,000 7,500 8,000 8,500 9,000 9,500 10,000
500
3,500
3,000
2,500
2,000
1,500
1,000
40
500
1,000
DLS
(a)
1,000
3,000
3,500
4,000
4,500
5,000
5
4,350
Side View
4,000
MD (ft)
4,500
5,000
5,500
4,450
4,500
4,550
4,600
4,650
MD (ft)
5
0
0
3,500
4,400
Top View
5
4,350
3,000
1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000 10,000
MD (ft)
5,500
500
0
2,500
0
0
(b)
10
0
2,500
(c)
Projections, (in)
60
0
4,400
50
4,450
4,500
100
4,550
150
4,600
200
(d)
MD (ft)
4,650
250
CF (lbf)
Fig. 5Sliding operation at TD, Field Case 1: (a) axial load, (b) contact-force distribution, (c) close-up of contact force in vertical
section, and (d) helically buckled drillstring position in wellbore. CF 5 coefficient of friction.
September 2014 SPE Drilling & Completion
ID: jaganm Time: 20:26 I Path: S:/3B2/DC##/Vol00000/140027/APPFile/SA-DC##140027
285
60
45
30
15
0
0
500
1,000
1,500
2,000
(c)
10
Axial Load (klbf)
2,500
DLS
(a)
DLS
12
60
45
30
15
0
0
1,000
Page: 286
Total Pages: 16
1,500
2,000
2,500
1,500
MD (ft)
2,000
2,500
3,000
14
7,000
16
18
6,000
(b)
DLS
22
0
60
45
30
15
0
0
500
1,000
500
1,000
1,500
MD (ft)
1,500
2,000
2,500
2,000
2,500
3,500
New Soft String
New Stiff String
3,000
2,500
20
Torque (lbft)
500
Stage:
5,000
4,000
3,000
2,000
2,000
1,500
1,000
1,000
500
0
0
0
500
1,000
1,500
MD (ft)
2,000
500
2,500
1,000
Fig. 6Drilling operation, Field Case 2: (a) axial load, (b) torque, and (c) contact-force distribution.
Table 5Field Case 2: T&D results comparison for hookload and surface torque.
286
Stage:
DLS
DLS
Total Pages: 16
20
20
10
0
0
2,000
4,000
6,000
8,000
10,000
10
0
0
12,000
16,000
3,000
14,000
2,500
Contact Force (lbf)
Torque (lbfft)
Page: 287
12,000
10,000
New Soft String
New Stiff String
Program A Soft String
Program A Stiff String
Program B Soft String
Program B Stiff String
8,000
6,000
2,000
4,000
4,000
6,000
8,000
6,000
8,000
10,000
12,000
1,500
1,000
500
4,000
0
2,000
2,000
6,000
8,000
10,000
12,000
0
0
2,000
4,000
MD (ft)
MD (ft)
(a)
(b)
10,000
12,000
Fig. 7Rotary drilling, Field Case 3: (a) torque and (b) contact-force distribution.
Table 7Field Case 3: T&D results comparison for hookload and surface torque.
287
Stage:
Page: 288
Total Pages: 16
results for surface T&D up/down are close enough for all six
models to be within the uncertainty range associated with the
commonly used, lumped-parameter friction factor. However,
some major differences in hookload for sliding and slackoff operations are observed, which are shown to be caused by differences
in location and magnitude of contact force between the drillstring
and wellbore. Further, significantly lower surface torque is predicted by the new dynamic stiff string compared with other models for one case history because of lower contact forces in the
vertical section of the well.
The key finding of this paper is that major differences are
observed for contact forcesboth location and magnitudefor
the new dynamic stiff-string model compared with all five other
models, including the two static stiff-string models. These differences in contact forces are most significant when the drillstring
has helically buckled or when doglegs in the wellbore are high.
Helical buckling of the drillstring can harm drilling-optimization
efforts by preventing effective weight transfer to the bit, thereby
reducing rate of penetration or causing poor motor performance.
The difference in contact-force prediction would also be important for prediction of fatigue of a rotated drillstring and for prediction of casing and drillpipe wear.
Comparison of New Stiff- and Soft-String Models Against
Field Data for Field Case Histories 3 and 4. As discussed in the
preceding section, the new stiff-string models compared favorably
in most situations against well-accepted industry Programs A and
B. This good agreement provides some level of validation for the
new stiff-string model. The next step in validation of the new stiffstring model is to compare its predictions against field data.
In this section, the new stiff-string model is compared against
field data for the same Field Case History 3 described in the preceding section and against field data for a fourth well, which is
denoted Field Case History 4. Data from Field Case History 3 are
depth-based data taken from daily drilling reports (Table 8),
whereas data from Field Case History 4 are time-based. Depthbased and time-based field data have different advantages and disadvantages. Both types are used as part of the validation process
to illustrate the viability of the new stiff-string model. For these
comparisons, the new soft-string-model predictions will also be
included to illustrate potential differences with the new stiff-string
model. As noted earlier, the new soft-string model agrees very
closely with industry-accepted soft-string Models A and B.
Depth-Based Field Data From Field Case History 3. Depthbased data were obtained from daily reports for the Field Case His-
Stage:
Page: 289
Total Pages: 16
250
160
140
100
Field Data
Field Trend
New Stiff String
New Soft String
80
60
40
200
120
150
Field Data
Field Trend
New Stiff String
New Soft String
100
50
20
0
0
0
2,000
4,000
6,000 8,000
MD (ft)
10,000
12,000 14,000
2,000
4,000
6,000
8,000
10,000
12,000 14,000
MD (ft)
data vs. the models in the horizontal section is not well understood, although the qualitative character of model predictions
appears to be correct in our view.
Comparing surface torque also helps validate the new stiffstring model against the field data. As shown in Fig. 11, the torque generally agrees with the new stiff-string-model predictions,
but contains two points of discrepancy in the deepest part of the
well in the horizontal section. These field data do not continue the
clear trend of increasing torque with increasing MD, which would
be expected and is predicted by both models. The likely conclusion is that the field data are compromised, and this clearly demonstrates the limitations of the use of only depth-based data
derived from daily reports. Table 11 shows that the new stiffstring-model predictions on average are 20% different from field
data, even including the two spurious points.
Table 11Average of absolute value of differences between field data and model predictions.
September 2014 SPE Drilling & Completion
ID: jaganm Time: 20:26 I Path: S:/3B2/DC##/Vol00000/140027/APPFile/SA-DC##140027
289
140
14
120
12
100
10
Torque (klb-ft)
80
Field Data
Field Trend
New Stiff String
New Soft String
60
40
Page: 290
Total Pages: 16
Field Data
Field Trend
New Stiff String
New Soft String
8
6
4
20
0
Stage:
2
2,000
4,000
6,000
8,000 10,000 12,000 14,000
MD (ft)
Fig. 10Field Case History 3: Comparisons of slackoff hookload for new stiff-string and new soft-string models vs. field
data.
0
0
2,000
4,000
6,000
8,000
MD (ft)
10,000
12,000 14,000
to total depth of approximately 19,000 ft. Some of the other parameters that were used for this well are shown in Table 12 and
the drillstring design is listed in Table 13.
The following graphs show the results obtained by running the
new stiff- and soft-string models and comparing them with the
real-time data obtained from the field. This case also shows good
agreement with the predictions, although some of the data have
significant of noise.
Rotating off bottom (ROB) was determined by selecting rev/
min between 50 and 100; operation tag of Off Bottom; and rate
of penetration between 5 and 5 ft/min. Rotary drilling was determined by selecting weight on bit (WOB) greater than 10 klbf; rev/
min of greater than zero; and the operation tag of On Bottom.
Tripping-in and tripping-out operations were determined by use
of the Off Bottom tag as true and WOB less than 5 klbf. For
tripping in, speeds of greater than 100 ft/hr were selected, and for
350
300
250
200
Field Data
New Soft String
New Stiff String
150
100
50
5,000
10,000
15,000
20,000
MD (ft)
Fig. 12Field Case History 4: Comparisons of ROB hookload
for new stiff-string and new soft-string models vs. field data.
290
tripping out, speeds of less than 100 ft/hr were used. In addition,
the difference in the bit position was checked, and for tripping in
was made sure to be positive (and negative for the tripping-out
operation).
Again, the first step in comparison is to examine ROB to
ensure predicted buoyed weight of the string matches field data
with acceptable accuracy. Results are shown in Fig. 12, which
shows the available field data were limited to the horizontal section of the well. Predictions from the new stiff-string model agree
closely with the lower limit of the filtered time-based data, which
indicates a good match.
After matching the ROB hookload, the next step is to calibrate
the friction factors by comparing the pickup and slackoff hookloads. The best agreement was obtained with a 0.17 openhole friction factor and a 0.20 cased-hole friction factor. Both of these are
within the range of common friction factors observed in other
wells for drilling operations. Results are shown in Fig. 13a for
slackoff hookload and Fig. 13b for pickup hookload.
The slackoff data in Fig. 13a were obtained partly during a trip
and partly during connections. Moreover, data were only available
for portions of the well. Data from approximately 9,00014,000-ft
measured depth represent a trip and provide less spread compared
with connection data. Similarly for pickup data in Fig. 13b, data
from approximately 14,000 ft to surface represent a trip, and again
show a smaller spread after filtering compared with the connection data. Overall, the agreement is within acceptable limits.
The next step is to compare surface torque while drilling and
ROB. By use of an estimated bit torque of 6,000 lbf-ft and the
same friction factors as discussed previously, comparisons are
shown for torque while drilling (Fig. 14a) and torque while ROB
(Fig. 14b). The method for filtering the torque while drilling was
to select all the torques less than 5 klbf-ft. For ROB, all torques
with values of zero were taken out because this would suggest no
rotation. Then, for both drilling and ROB, the torques for both
September 2014 SPE Drilling & Completion
Stage:
450
400
Field Data
New Soft String
New Stiff String
350
300
Total Pages: 16
450
400
250
200
150
100
50
Page: 291
5,000
10,000
MD (ft)
15,000
300
250
200
150
100
50
20,000
Field Data
New Soft String
New Stiff String
350
5,000
(a)
10,000
MD (ft)
15,000
20,000
(b)
Fig. 13Field Case History 4: Comparisons of (a) slackoff and (b) pickup hookload for new stiff-string and new soft-string models
vs. field data.
20
Field Data
New Soft String
New Stiff String
15
20
10
10
0
0
5,000
10,000
15,000
20,000
Field Data
New Soft String
New Stiff String
15
5,000
10,000
Depth (ft)
Depth (ft)
(a)
(b)
15,000
20,000
Fig. 14Field Case History 4: Comparisons of (a) surface torque while drilling and (b) surface torque while ROB for new stiff-string
and new soft-string models vs. field data.
were arranged in order of depth, and all values for the same depth
were averaged to determine a mean value.
The comparison of torque while drilling shows the predictions
for the new stiff-string model falling near the center of the cloud
of time-based field data, which indicates good agreement. It
should be noted that moderate stick/slip occurred during much of
the time represented by these data, which is indicated by the
spread in field data. The torque-while-ROB comparison shows a
similar result to the torque-while-drilling comparison. Again, the
new stiff-string-model predictions lie near the center of the cloud
of field data, which represents acceptable agreement.
In summary, predictions from the new stiff-string model show
acceptable agreement with time-based field data from Field Case
History 4. Combining this with previously noted good agreement
with depth-based field data from Field Case History 3 and with
comparisons against industry-accepted Programs A and B, the
new stiff-string model has been shown to provide suitably accurate predictions for all cases tested so far.
Future Work
The dynamic stiff-string model provides improved capability to
predict local contact force and stress, which we plan to use to
study drillstring and casing wear and drillstring fatigue. Additional field validation is considered important and is ongoing. The
new model will also be used to study noncircular wellbore cross
sections, which our initial dynamic simulations indicate have im-
291
R
s
t
T
v
vh
vr
w
y
z
a
d
d0
DL
Dt
h
j
j1
k
l
qIp
w
X
Stage:
Page: 292
Total Pages: 16
Acknowledgments
We gratefully acknowledge major contributions from Lee Winchell for his technical guidance and expertise and for providing
field case histories; Michael Kuhlman for his technical guidance
and expertise on field practices and providing field case histories;
Anand Kannappan, Melissa Frilot, and Mike Rossing for their
technical guidance and expertise; and Collin Johnson for his
expert technical writing. We also appreciate Weatherfords support in conducting this work and publishing this paper.
Nomenclature
A cross-sectional area of the drillstring
AE axial stiffness of the drillstring
D outer diameter of the drillstring
Dh hole diameter
E Youngs modulus
EI bending stiffness of the drillstring
fn complex external forces per unit length
fx projection of the external force per unit length onto axis
oe1
fy projection of the external force per unit length onto axis
oe2
fz projection of the external force per unit length onto axis
oe3
G shear modulus
GIp torsional stiffness of the drillstring
I axial moment of inertia of the drillstring
Ip polar moment of inertia of the drillstring
j (1)1/2 imaginary unit
kh friction factors in the circumferential direction
kx friction factors in the axial direction
L drillstring length
m mass of the drillstring
m* effective mass, which includes mass of the drillstring m
and the added fluid mass inside the drillstring and in the
annulus per unit length
M torque
Mb bending moment
p contact force per unit length
P weight on bit
Q torque on bit
r complex lateral displacement of drillstring axis about
wellbore axis
Aadnoy, B.S., Cooper, I., Miska, S.Z., et al. 2009. Advanced Drilling and
Well Technology. Richardson, Texas: SPE.
Aslaksen, H., Anand, M., Duncan, R., et al. 2006. Integrated FEA Modeling Offers System Approach to Drillstring Optimization. Presented at
IADC/SPE Drilling Conference, Miami, Florida, 2123 February.
SPE-99018-MS. http://dx.doi.org/10.2118/99018-MS.
Cernocky, E.P. and Scholibo, F.C. 1995. Approach to Casing Design for
Service in Compacting Reservoirs. Presented at SPE Annual Technical
Conference and Exhibition, Dallas, Texas, 2225 October. SPE30522-MS. http://dx.doi.org/10.2118/30522-MS.
Cernocky, E.P. and Scholibo, F.C. 1994. Approach to Using Abaqus to
Simulate Drilling and Casing Wells and Loading Casing Through Reservoir Compaction. Proc., Seventh Abaqus Users Conference, Providence, Rhode Island, 13 June, Sec. Contributed Papers, 167176.
Glowka, D.A. 1989. Use of Single-Cutter Data in the Analysis of PDC Bit
Designs: Part 1 Development of a PDC Cutting Force Model. J Pet Technol 41 (8): 797849. SPE-15619-PA. http://dx.doi.org/10.2118/15619-PA.
Ho, H.-S. 1986. General Formulation of Drillstring Under Large Deformation and Its Use in BHA Analysis. Presented at SPE Annual Technical
Conference and Exhibition, New Orleans, Louisiana, 58 October.
SPE-15562-MS. http://dx.doi.org/10.2118/15562-MS.
Ho, H.-S. 1988. An Improved Modeling Program for Computing the Torque and Drag in Directional and Deep Wells. Presented at SPE Annual
Technical Conference and Exhibition, Houston, Texas, 25 October.
SPE-18047-MS. http://dx.doi.org/10.2118/18047-MS.
Johancsik, C.A., Friesen, D.B., and Dawson, R. 1984. Torque and Drag in
Directional Wells Prediction and Measurement. J Pet Technol 36
(6): 987992. SPE-11380-PA. http://dx.doi.org/10.2118/11380-PA.
Mason, C.J. and Chen, D. C.-K. 2007. Step Changes Needed To Modernise T&D Software. Presented at SPE/IADC Drilling Conference, Amsterdam, The Netherlands, 2022 February. SPE-104609-MS. http://
dx.doi.org/10.2118/104609-MS.
292
contact force
arc length
time
axial force
resultant velocity of a point on the drillpipe surface
circumferential velocity of a point on the drillpipe
surface
axial speed, running speed, or rate of penetration
weight of the drillstring length in the mud per unit length
projections of the drillstring axial line on axis oe2
projections of the drillstring axial line on axis oe3
inclination angle of the wellbore
penetration of drillstring into the wellbore wall or deformation of the wellbore wall
clearance between the wellbore wall and the drillstring
mesh size
integration timestep
azimuth angle of the wellbore
projection of the well curvature onto plane oe2e3
projection of the well curvature onto axis oe1
stiffness of the borehole wall per unit length as defined
by the elastic characteristic of the formation
Frictional torque per unit length
moment of inertia of the drillstring
torsion angle
drillstring angular velocity, rev/min
References
McSpadden, A.R., Coker, O.D., and Ruan, G.C. 2011. Advanced Casing
Design with Finite-Element of Effective Dogleg Severity, Radial Displacements, and Bending Loads. Presented at PE Production and Operations Symposium, Oklahoma City, Oklahoma, 2729 March. SPE141458-MS. http://dx.doi.org/10.2118/141458-MS.
Menand, S., Sellami, H., Tijani, M., et al. 2006. Advancements in 3D
Drillstring Mechanics: From the Bit to the Topdrive. Presented at the
SPE/IADC Drilling Conference, Miami, Florida, 2123 February.
SPE-98965-MS. http://dx.doi.org/10.2118/98965-MS.
Mitchell, R.F. 2008. Drillstring Solutions Improve the Torque-Drag Model.
Presented at the IADC/SPE Drilling Conference, Orlando, Florida, 46
March. SPE-112623-MS. http://dx.doi.org/10.2118/112623-MS.
Mitchell, R.F. and Miska, S.Z. 2011. Fundamentals of Drilling Engineering, Vol. 12. Richardson, Texas: Textbook Series, Society of Petroleum Engineers.
Mitchell, R.F. and Samuel, G.R. 2007. How Good is the Torque-Drag
Model? Presented at the IADC/SPE Drilling Conference, Amsterdam,
The Netherlands, 2022 February. SPE-105068-MS. http://dx.doi.org/
10.2118/105068-MS.
Newman, K.R., Corrigan, M., and Cheatham, J.B. 1989. Safely Exceeding
The Critical Buckling Load In Highly Deviated Holes. Paper presented at Offshore Europe, Aberdeen, United Kingdom, 58 September. SPE-19229-MS. http://dx.doi.org/10.2118/19229-MS.
Newman, K.R. and Proctor, R. 2009. Analysis of Hook Load Forces During Jarring. Presented at SPE/IADC Drilling Conference, Amsterdam,
The Netherlands, 1719 March. SPE-118435-MS. http://dx.doi.org/
10.2118/118435-MS.
Rezmer-Cooper, I., Chau, M., Hendricks, A., et al. 1999. Field Data Supports the Use of Stiffness and Tortuosity in Solving Complex Well
Design Problems. Presented at SPE/IADC Drilling Conference, Amsterdam, The Netherlands, 911 March. SPE-52819-MS. http://
dx.doi.org/10.2118/52819-MS.
Sheppard, M.C., Wick, C., and Burgess, T. 1987. Designing Well Paths to
Reduce Torque and Drag. SPE Drill Eng 2 (4): 344350. SPE-15463PA. http://dx.doi.org/10.2118/15463-PA.
Smalley, E. and Newman, K. 2005. Modeling and Measuring Dynamic
Well Intervention Stack Stress. Presented at the SPE/ICoTA
Coiled Tubing and Well Intervention Conference, The Woodlands,
Texas, 1213 April. SPE-94233-MS. http://dx.doi.org/10.2118/94233MS.
Tikhonov, V.S., Safronov, A.I., and Gelfgat, M.Ya. 2006. Method of
Dynamic Analysis of Rod-in-Hole Buckling. Proc., 8th Biennial ASME
Conference on Engineering Systems Design and Analysis, Torino,
Stage:
Page: 293
Total Pages: 16
293
Albert Nurgaleev is Lead Engineer Technical Support for Tubular and Premium Connections at TMK Company. Previously he
worked for 6 years in Research & Development with Weatherford in Russia. Nurgaleevs current interests include stress analysis, fatigue tests, and torque-and-drag analysis. He holds a
bachelors degree in physics engineering from the University
of Russia. Nurgaleev is a member of SPE.
Lev Ring has been the Director of Technology Development
with Weatherford since 2002. His current interests include
managed-pressure drilling, drilling automation, and real-time
simulation of drilling processes. Ring holds a PhD degree in
physics and mathematics from the Russian Academy of
Sciences and a masters degree in applied physics from the
Institute of Physics and Technology, Russia. He is a member of
SPE.
Varadaraju Gandikota is Manager of Secure Drilling for Deepwater Systems Engineering with Weatherford. He has been
with the company for more than 8 years and has been
involved in numerical modeling and simulation for the past 15
years. Gandikota previously worked in other industries in
applied research, engineering, and numerical modeling. His
current interests include managed-pressure drilling, drilling-systems modeling, computational fluid dynamics, and fluid/structure interaction. Gandikota holds a masters degree in
294
Stage:
Page: 294
Total Pages: 16