You are on page 1of 7

Effects of Well Deviation

on Helical Buckling
R.F. Mitchell,SPE, Enertech Engineering and Research Co.

Summary tact loads. An expanded discussion of this simplified analysis with


Current helical buckling models are valid for vertical wells, but the practical applications has been presented as a separate paper.7
validity of these solutions for deviated wells was not known. This
paper describes the numerical solution of the nonlinear buckling Buckling Models for Deviated Wells
equations for arbitrary well deviation. Stability criteria are devel- Analysis of Helical Buckling. This section describes the theoretical
oped for lateral and helical buckling, and simple correlations are de- basis for the analysis of buckling in deviated wellbores. The first
veloped for buckling length change, maximum bending stress, and generally useful buckling solution was published by Lubinski.1 In
contact force. Lubinski’s analysis, the wellbore is assumed to be vertical and the
tubing buckled shape was modeled as a helix with variable pitch.
Introduction Mitchell4 showed that Lubinski’s solution was an approximate solu-
The most generally accepted method for the analysis of buckling, tion to the beam-column equations8 with displacements constrained
tubing movement, and packer selection is the method developed by to a cylinder. In this formulation of the beam-column equations, the
Lubinski et al.1. Analyses following Lubinski’s basic approach lateral displacements, shown in Fig. 3, are given by
have been developed for more complicated tubing configuratioins, u 1 + r cos q . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (1)
e.g. tapered strings.2,3. Henry Woods, in the Appendix to Ref. 1, de-
veloped a mechanical model of well-buckling behavior that pre-
dicted the buckled configuration as a function of well loads. The and u 2 + r sin q. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (2)
model featured the following. where q is the helix angle, and r is the tubing/casing radial clearance.
1. The slender-beam theory was used to relate bending moment The differential equation for the helix angle q is given by
to curvature.
* EIqȀȀȀȀ ) ƪ2EI(qȀ) * FqȀƫȀ ) ǒ w lńr Ǔ sin q + 0,
3
2. The tubing was assumed to buckle into a helical shape. . . . . (3)
3. The wellbore was assumed to be straight and vertical.
4. The pitch of the helix was related to the buckling load through where wl is the lateral tubing weight per unit length, EI is the bend-
the principle of virtual work. ing stiffness, F is the axial buckling force, and Ȁ denotes d/dz. The
5. Friction between the tubing and casing was neglected. axial buckling force F and the lateral tubing weight are both strongly
Mitchell developed a more general approach replacing the virtual influenced by fluid pressures and must be formulated accordingly.9
work relations with the full set of beam-column equations Lubinski’s helical pitch solution
constrained to be in contact with the casing.4 Helical buckling in a
deviated well, in this formulation, is described by a second order qȀ +" Ǹ2EI
F . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (4)
non-linear differential equation. For a vertical well, the solution to
this equation can be accurately approximated by the simple algebra- is an approximate solution to Eq. 3 for wl equal 0, but not the only
ic equation discovered by Lubinski and Woods. Because of the later- possible solution.
al gravity forces, it is not clear if this solution is valid for deviated The quantity qȀ is important and will appear often in the following
or horizontal wells. analysis. It can be related to the more familiar quantity pitch P
This paper presents a numerical solution to the buckling differen- through
tial equation. The Galerkin5 technique is applied to the equation us-
ing cubic interpolation functions. The resulting nonlinear algebraic P + 2pńqȀ. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (5)
equation is solved using Powell’s method.6 Calculation of results, Other important quantities, such as pipe curvature, bending mo-
including buckling length change, tubing contact forces, and dogleg ment, bending stress, and tubing length change are proportional to
angle are developed. As a test, the solution technique duplicated Lu- the square of qȀ. Nonzero qȀ indicates that the pipe is curving, while
binski’s solution for a vertical well, except near the neutral point, zero qȀ indicates that the pipe is straight.
where the Lubinski solution does not satisfy the buckling equation. An approximate solution to Eq. 3 for inclined wells, with F
This paper solves several sample problems and examines the effects constant, is
of well deviation on stability.
Two types of buckling behavior were found, with the particular q + a sinǒ bz Ǔ, a Ơ 1, . . . . . . . . . . . . . . . . . . . . . . . . . . . . (6)
form determined by the magnitude of the axial load. For small val-
ues of the axial force, lateral buckling or “snaking” was developed. which requires
A sketch of this type of buckling is shown in Fig. 1. When the axial
force exceeds a critical force determined by the well deviation, later- F y Ǹ4w lEIńr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (7)
al buckling transitions into full helical buckling, as sketched in Fig. for b to be real. This is the Paslay-Dawson equation, derived in a dif-
2. The results in this paper characterize the two forms of buckling ferent way.10
and determine the critical loads associated with each buckling form. A dimensionless form of Eq. 3 is developed to minimize the num-
Because of the intense numerical calculations needed to solve the ber of parameters needed to develop correlations. The Woods char-
deviated well buckling problem, the results of the analysis were acteristic length l is defined
used to develop simple analytical expressions for use in tubing anal-
ysis, e.g. buckled length change, maximum bending stress, and con-
l+ Ǹ2EI
F
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (8)
o
Copyright 1997 Society of Petroleum Engineers

Original SPE manuscript received for review 8 May 1995. Revised manuscript received 14 Au-
with the dimensionless length c defined
gust 1996. Paper peer approved 11 December 1996. Paper (SPE 29462) first presented at the
1995 Production Operations Symposium held in Oklahoma City, Oklahoma, 2–4 April 1995. c + zńl. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (9)

SPE Drilling & Completion, March 1997 63


Fig. 1—Lateral “snake” buckling.
Fig. 2—Helical buckling.

Eq. 3 written in this dimensionless form becomes


and w a + w cos f . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (13)

dc
ƪ ǒ N Lu Pa
Ǔƫ
d * q.. ) 2 q. 2 * 1 ) c q. ) N 2 sin q + 0 . . . . . . (10)
where w is the weight per unit length as given by Lubinski1 in his
equation 5. The Paslay number can be rewritten
where
Ǹ4wEIńr
F N Pa + Ǹsin f. . . . . . . . . . . . . . . . . . . . . . . . . . . (14)
N Lu + o + Lubinski Number, . . . . . . . . . . . . . . . . . . . (11) Fo
wa l
Note that the effects of deviation include not only the angle from
the vertical f, but also the bending stiffness of the tubing, EI, and
Ǹ4w lEIńr the radial clearance r. For a fixed deviation angle f and fixed hole
N Pa + + Paslay Number, . . . . . . . . . . . . . . . . (12) diameter, buckling behavior would be significantly different, for ex-
Fo
ample, for coiled tubing compared to conventional drill pipe, since
wa is the axial weight per length, and “@” denotes d/dc. Notice that the bending stiffness and radial clearances would be very different.
Eq. 10 has only two dimensionless parameters, NLu and NPa, which Eq. 10 was solved numerically. This analysis is described in detail
influence the solution. The Lubinski number, NLu, is the dimension- in Appendix A. A test case with NPa equal to zero was executed, with
less buckled length of the tubing. Because the characteristic length results shown in Fig. 4. Notice that the dimensionless force F/Fo
is proportional to the pitch of the helix analyzed by Lubinski, the Lu- was used as the axial variable. This facilitates comparison of curves
binski number also roughly represents the number of “coils” in a with different values of NLu, where c coordinates would not corre-
buckled helix. In the example presented by Lubinski,1 NLu equals spond. The numerical solution matches the Lubinski solution very
2,714, though this is somewhat of an extreme case. well, except near the neutral point (since we are dealing with the
The Paslay number, NPa, is the ratio of the Paslay-Dawson buck- buckling force, this neutral point includes the effects of internal and
ling stability force10 to the applied force Fo . For a Paslay number external pressure). An expanded view of the neutral point is shown
greater than 1, buckling is suppressed. For a well deviated from the in Fig. 5. The Lubinski solution was derived in Ref. 4 by neglecting
vertical by angle f, the axial and lateral weight per unit lengths can .

be expressed derivatives of q , which can be seen to be large (a nearly vertical


curve) for the Lubinski solution near the neutral point in Fig. 5. Here
w l + w sin f the numerical solution appears more reasonable.

64 SPE Drilling & Completion, March 1997


Fig. 4—Comparison of numerical and Lubinski solution.

Fig. 3—Coordinate system for buckling analysis.

Example Calculations of Buckling in Inclined Wellbores. Figs.


6 and 7 show the results of buckling in an inclined wellbore with
NPa= 0.45. For this case there is no helical buckling, only lateral or
“snake” buckling. In Fig. 6, the variation of q with dimensionless Fig. 5—Comparison of numerical and Lubinski solution near the
force is plotted. Notice that q is less than p/2, i.e. the tubing only neutral point.
rides halfway up the sides of the casing. Notice also that for dimen- .
sionless force less than 0.45 there is still some buckling. Based on increases q as the helix descends the casing wall. The tubing stays
the Paslay-Dawson criterion, buckling should be suppressed for the in lateral buckling when the dimensionless force falls below about
local Paslay number greater than 1. In this case this corresponds to 0.15, and buckling is suppressed for force below 0.05 .
a dimensionless force F/Fo t0.45. This extra buckling is a bound-
ary effect caused by the buckling
. for F/Fo u0.45. As shown in Fig. Transition from Lateral Buckling to Helical Buckling. Figs. 9
7, the maximum value of q is less than the NPa+0.0 curve. This and 10 show the minimum NPa for which helical buckling exists for
means that the curvature, and the resulting bending stress, is less for two different values of NLu. In both cases the minimum value is
the lateral buckling result than for the helical buckling result.
. Since approximately 0.35, suggesting that this minimum value is indepen-
buckling length change is proportional to the square of q , buckling dent of NLu, at least for NLuu200. According to Chen et al.11, heli-
length change will also be less than the helical buckling case. For cal buckling transition should occur at NPa equal to 1/Ǹ2^0.707.
fixed packers, this implies higher axial loads relative to pure helical This result is based on an energy balance calculation, which sug-
buckling. For free packers, this means that seal lengths will be short- gests that in the present calculation, there is enough energy in the
er than for pure helical buckling.. string to form a helix, but not enough force to lift the string into the
Fig. 8 shows the results for q in the case of a small inclination helical configuration. To test this theory, the model was run with the
from vertical. In this case we see combined helical and lateral buck- wellbore initially vertical, then lateral forces were applied. The re-
ling. Small variations in the helical
. buckling curve are due to the lat- sult, Fig. 11, shows buckling transitions from helical to lateral at
eral forces which hold back q as the helix climbs the casing wall and higher Paslay numbers. For example, for NPa = 0.25, transition from

.
Fig. 6—Helix angle for buckling in inclined wells. Fig. 7— q fog buckling in inclined wellbores.

SPE Drilling & Completion, March 1997 65


Fig. 8—Slight inclination; small Paslay number. Fig. 9—Transition from lateral to helical buckling.

Fig. 10—Transition from lateral to helical buckling. Fig. 11—Transition from helical to lateral buckling/unloading.

helix to lateral buckling occurs at about the local Paslay number of . 0.460 0.040
q max + 1.12271ǒf * N PaǓ f , . . . . . . . . . . . . . . . . . (16)
about 0.75, where the local Paslay number is defined
where f is F/Fo , the dimensionless force used in the figures. Eq. 16
Ǹ4w lEIńr applies only for f greater than NPa . Figs. 12 through 15 show this
N Pal + , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (15)
F to be a good correlation for a variety of NPa and NLu.
and F is evaluated at the current point rather than at the boundary
Maximum Bending Stress. The bending stress sb in a slender
condition.
beam, distance y from the centroid, is given by
The model results suggest the following conclusions:
1. No buckling occurs for local Paslay Numbers greater than one. s b + MyńI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (17)
2. Lateral buckling occurs for 1yNPay1/Ǹ2.
where the bending moment M is
3. For loading, lateral buckling transitions to helical buckling for
the local Paslay number less than 0.35, roughly ½ the Chen result. M + EIrqȀ 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (18)
4. For unloading, helical buckling transitions to lateral buckling
The maximum value of the bending stress occurs at the outer fiber
for the local Paslay number y1/Ǹ2, the Chen result.
of the tubing, i.e. y equals ½ the outside diameter do
These results are summarized in Table 1 in terms of the reciprocal
of the Paslay Number, so that they will be more directly comparable Erd o . 2
s b max + q . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (19)
to previous analyses.10,11 2 l 2 max

Amplitude of Lateral Buckling. The amplitude of the lateral buck- Buckling Length Change. In general, the local buckling strain eB
ling is needed for calculating the maximum bending stress caused is defined
by buckling and the buckling length change. The lateral buckling
e B + ½r 2qȀ 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (20)
amplitudes calculated by the buckling analysis were least-squares
curve fitted to give the correlation and the buckled length change is the integral of Eq. 20 over the

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
buckled length. In the lateral buckled section, the local buckling
strain is not a useful concept because of the harmonic nature of the

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
TABLE 1—STABILITY OF HELICAL BUCKLING
buckling. Of more use is the average buckled strain related to the lat-

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
N *1
Pal t 1
No Buckling (Pasley) eral buckling amplitude already determined. Fig. 16 shows a typical
half-cycle of lateral buckling. A half-cycle sine wave is superim-

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
1 t N *1 Ǹ Lateral Buckling (Chen) posed to show that the buckled solution is not a simple harmonic, but
Pal t 2

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
is somewhat “fatter.” Numerical integration of the half-cycle gives
Ǹ2 t N *1 t 2 Ǹ2 Lateral or Helical Buckling
the average buckling strain to be

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
Pal

2 Ǹ2 t N *1 Helical Buckling Only 2 .2


Pal
e Bavg + 0.2890ǒ rńl Ǔ q max. . . . . . . . . . . . . . . . . . . . . . . . . (21)
Ǹ
66 SPE Drilling & Completion, March 1997
Fig. 12—Correlation of lateral buckling amplitude. Fig. 13—Correlation of lateral buckling amplitude.

Fig. 14—Correlation of lateral buckling amplitude. Fig. 15—Correlation of lateral buckling amplitude.

Fig. 16—Half cycle of lateral buckling. Fig. 17—Helix angle correlation.

Tubing Contact Force. The tubing/casing contact force is given by4 This correlation is illustrated
.
in Fig. 17.
Because both q and q vary, the average value of Ncf over a cycle
w n + w l cos q * rEIqȀ 4 ) rF qȀ 2, . . . . . . . . . . . . . . . . . . . (22) is more useful. The average value of cosq is:
which can be written in a dimensionless form as p

.4
N cf + N 2Pa cos q * q ) f q
.2
. . . . . . . . . . . . . . . . . . . . . . . (23)
(cos q) avg + 1ńp ŕ cos(q max sin x)dx + J o(q max) . . . . . . . (26)
0
where the dimensionless contact force Ncf is given by
where Jo is the zero-th order Bessel function. A polynomial repre-
.
4EIw n sentation of Jo can be found in Ref. 12. The average value of q was
N cf + . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (24)
rF 2o determined for the buckling length change:
.2 .2
For lateral buckling, q varies approximately sinusoidally with q avg + 0.5778q max . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (27)
amplitude qmax, that has been correlated .
The average value of q was also determined by numerical in-
0.57429
q max + 2.18409ǒ1 * N PalǓ for N Pal u .70 tegration:
.4 .4
q max + 0.39412 ) 2.33584ǒ1 * N PalǓ for N Pal t .70. . . (25) q avg + 0.4590q max . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (28)

SPE Drilling & Completion, March 1997 67


ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ Gi + residual error of ith equation

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
TABLE 2—BUCKLING CRITERIA FOR SAMPLE PROBLEM
I+ moment of inertia

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
Paslay Buckling Helical Buckling L+ the tubular string length
g f
Deviation Angle force Force

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
M+ bending moment
(degrees) (lbf) (lbf)
Ncf + dimensionless contact force

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
0 0 0 NLu + Lubinski number+Fo /(wa l)
NPa +

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
1 1,158 3,275 Paslay number+sqrt(4wl EI/r)/Fo
2 1,637 4,630 NPal + local Paslay number+sqrt(4wl EI/r)/F

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
5 2,587 7,317 r+ the tubing/casing radial clearance
u1+

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
10 3,652 10,329 displacement in the x-coordinate direction
20 5,125 14,496 u2+ displacement in the y-coordinate direction

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
30 6,197 17,528 w+ Lubinski equivalent weight per unit length

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
40 7,026 19,873 wa + axial distributed load in the tubing
wl + lateral distributed load in the tubing

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
50 7,670 21,694
60 8,155 23,066 wn + the contact force between the tubing and casing

ÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁÁ
70 8,495 24,027 Xi + ith natural boundary condition
90 8,763 24,786 y+ distance from centroid
z+ measured depth (Caution: in this development, z is
measured from the surface. In some papers, z is
Eq. 23 can be evaluated for an average value of Ncf through the
measured from the bottom of the string.)
use of Eqs. 27 through 29.
a+ amplitude of approximate solution
Sample Calculation. The example problem in Ref. 1 consisted of b+ coefficient in approximate solution
a 27/8-in. 6.5-lbm/ft tubing buckled inside a 7 in. 32 lbm/ft-casing. l+ Lubinski characteristic length
For this case, w+0.640 lbf/in., I+1.61 in.4, and r+1.61 in. sb + bending stress
Using these values, we can evaluate the buckling criteria for vari- sbmax+ maximum value of bending stress
ous angles of the deviation angle f. Table 2 gives the Paslay buck- Fn + nth cubic interpolation function
ling force and the helical buckling transition force as a function of f+ deviation angle measured from the vertical
the deviation angle. The first observation is that the buckling forces q+ angle between the pipe center location and the
for even very small deviations are surprisingly high. This suggests x-coordinate axis
.
that the conventional analysis may start to loose accuracy at surpris- q max+ lateral buckling amplitude correlation
ingly small deviation angles. A second observation is that the buck- c+ dimensionless length
ling forces are insensitive to angles above about 50°. Most of the ef-
fects of transition between vertical wells and deviated wells occur References
in the range of 5 to 45°. Finally, notice how large the helical buckling 1. Lubinski, A., Althouse, W.S., and Logan, J.L.: “Helical Buckling of
transition forces become for angles greater than about 10°. This sug- Tubing Sealed in Packers,” JPT (June 1962) 655.
gests that helical buckling may be surpressed in deviated wells, ex- 2. Hammerlindl, D.J.: “Movement, Forces, and Stresses Associated With
cept in very severe cases. Combination Tubing Strings Sealed in Packers,“ JPT (February 1977)
195; Trans., AIME, 273.
Results and Conclusions 3. Hammerlindl, D.J.: “Packer-to-Tubing Forces for Intermediate Pack-
ers,” JPT (March 1980) 195-208; 515.
Accurate solution of the buckling equations is important for several 4. Mitchell, R.F.: “New Concepts for Helical Buckling,” SPEDE (Septem-
reasons. Bending stresses due to tubing buckling will be overesti- ber 1988), 303.
mated for deviated wells using Lubinski’s formula. However, Lu- 5. Finlayson, B.A.: The Method of Weighted Residuals and Variational
binski’s solution applied to deviated wells will also overpredict tub- Principles, Academic Press, New York, (1972).
ing movement. For a fixed packer, this solution will overestimate 6. Powell, M.J.D.: “A Hybrid Method for Nonlinear Equations,“ Numeri-
tubing compliance, which may greatly underestimate the axial cal Methods for Nonlinear Algebraic Equations, P. Rabinowitz (ed.),
loads, resulting in a nonconservative design. For a free packer or Gordon and Breach, London (1970).
polished bore receptacle, exaggerated tubing motion will require 7. Mitchell, R.F.: “Buckling Analysis in Deviated Wells: A Practical Meth-
excessive seal length. Further, because tubing incremental motion od,” paper SPE 36761 presented at the 1996 SPE Annual Conference
will control the friction load direction, errors in overall tubing dis- and Exposition, Denver, Colorado, 6–9 October.
placement will generate further errors in friction loads. 8. Timoshenko, S.P. and Goodier, J.N.: Theory of Elasticity, McGraw-Hill
Book Co, New York (1970).
The following results and conclusions were developed in this paper:
9. Mitchell, R.F.: “Forces on Curved Tubulars Caused to Fluid Flow,”
1. A numerical method for the solution of the helical buckling
SPEPF (February 1996), 30.
equation in deviated wells. 10. Dawson, Rapier and Paslay, P.R.: “Drillpipe Buckling in Inclined
2. Stability criteria for (a) the transition from straight pipe to later- Holes,” JPT (October 1984).
al buckling, (b) the transition from lateral buckling to helical buck- 11. Chen, Yu-Che, Lin, Yu-Hsu, and John B. Cheatham: “Tubing and Cas-
ling, and (c) the transition back to lateral buckling from helical ing Buckling in Horizontal Wells, JPT (February 1990), 140.
buckling upon unloading. 12. Aziz, K. and Settari, A.: Petroleum Reservoir Simulation, Elsevier Ap-
3. Special correlations for lateral buckling for bending stresses, plied Science Publishers, New York (1979).
contact forces, buckling length changes, and maximum dogleg
angles. AppendixĊNumerical Solution of the
Buckling Differential Equation
Nomenclature
The dimensionless buckling differential equation
do + tubing outside diameter
E+ Young’s modulus
eB + buckling strain dc
ƪ ǒ N Lu
Ǔƫ
d * q.. ) 2 q. 2 * 1 ) c q. ) N 2 sin q + 0 . . . . . (A-1)
Pa
eBavg+ average buckling strain
fi + ith interpolation function is solved numerically, using the Galerkin method.5 In the Galerkin
F+ the buckling force method, often called the finite element method, the angle q is
Fo + the packer to tubing buckling force approximated by

68 SPE Drilling & Completion, March 1997


q + Sc iF i(c) for c in ǒc 1, c 2Ǔ, . . . . . . . . . . . . . . . . . . . . . (A-2) This method
. generates a set of four algebraic equations in un-
knowns qi , qi where the equations are integrated numerically using
where ci are coefficients to be determined and Fi (c) are called inter- ten-point Gauss integration. The Xi that appear in the equations are
polation functions. In this case, cubic interpolation functions were called natural boundary conditions. When the equations for the
chosen so that a full cycle of q(2p) could be approximated with a whole problem are assembled, the X’s cancel out of the equations
minimum number of elements. In this case, except at the top and bottom ends of the string. These X’s are re-
. . moved by setting boundary conditions at the top and bottom of the
q(c) + q 1F 1 ) q 1F 2 ) q 2F 3 ) q 2F 4 for c Ů ǒc 1, c 2Ǔ , string, which are usually specified displacements. For M elements,
. . . . . . . . . . . . . . . . . . . . . (A-3) the resulting set of Gi ’s form a system of 2(M)1) nonlinear equa-
tions with 2(M)1) unknowns. Since no general method exists for
2
solving a system of nonlinear equations, a numerical method was
where F 1 + 0.25(x * 1) (x ) 2), necessary. A variation of Newton’s method, called Powell’s meth-
od10 was used to solve Eq. A-5. In this method, an initial solution
F 2 + 0.25(x * 1) (x ) 1)ǒc 2 * c 1Ǔ,
2
guess is corrected by calculating a correction vector. For small error,
2 the correction vector is calculated with Newton’s method,
F 3 + 0.25(x ) 1) (x * 2),
*1
F4 + 0.25(x ) 1) (x * 1)ǒc 2 * c 1Ǔ ,
2 u + * ƪēG(u
D³ ³ ³ƫ
o)ńēu
³
G(u o), . . . . . . . . . . . . . . . . . . (A–7)
.

and x + 2ǒc 2 * c 1Ǔńǒc 2 * c 1Ǔ * 1. where the vector ³ u is the vector of qi and qi coefficients, and where
ēG/ ē ³
u is block tridiagonal, which can be solved efficiently.13 For
. .
Notice that q(ci ) equals qi , and q (ci ) equals qi , i+1,2. In general, larger errors, the method of steepest descent is used. For intermedi-
Eq. A-3 does not solve Eq. A-1 exactly, however, the coefficients ate error, linear combinations of DĂ ³ u from Newton’s method and
can be chosen to minimize the error in the solution. A technique that steepest descent are used. Error limits for Newton’s method, steep-
integrates the error times a weighing function over the interval est descent, or combinations of methods were determined by trial
(c1*c2) and then chooses the coefficients to minimize this inte- and error for this problem.
grated error, is called a method of weighted residuals.5 In the Galer- A two point boundary condition was chosen. The condition at .
the
kin method, the interpolation functions are chosen as the weighing end of the pipe above the neutral point (axial tension) was set q +0.
functions, so the integrated error is For the condition at the tubing end with the applied buckling load,
set q equal to ǸF oń2E I , the Lubinski solution. Axial forces were ap-
.

Xi * NJ ƪ dc N Lu
ǒ
d * q.. ) 2 q. 2 * 1 ) c q. ) N 2 sin q
Pa
Ǔƫ Nj plied in 40 load increments, with the solution converged at each load
increment. The boundary condition at the applied load was also in-
cremented.
F i(c)dc + 0 i + 1, 4. . . . . . . . . . . . . . . . . . . . . . . . . . . (A-4)
SI Metric Conversion Factors
Eqs. A-4 can be integrated by parts to give ft 333 333* E*01 +m3

Xi * ŕ qF dc * ŕ 2ǒq * 1 ) Nc ǓqF dc ) ŕ N
.. ..
i
.2

Lu
. .
i Pa sin q
2
*Conversion factor is exact. SPEDC

Robert F. Mitchell is Vice President of Enertech Engineering and


F idc + G i + 0 i + 1, 4 . . . . . . . . . . . . . . . . . . . . . . . . . (A-5) Research Co. He has published over 40 papers on wellbore and
well completion problems, including thermal/flow simulation,
where drillstring mechanics, tubing buckling analysis, arctic well
completions, tubular stress analysis, and borehole stability. He
..
X1 + q1 * 2 q1 * 1 ) ǒ .2 c1 .
Ǔq,
N Lu 1
previously worked at Exxon Production Research Co., and holds
BA, MME, and PhD degrees from Rice U. He currently is a techniĆ
cal editor for SPE Journal. He is a registered professional engiĆ
.. neer in Texas.
X 2 + q 1,

..
X3 + * q2 ) 2 q2 * 1 ) ǒ .2 c2 .
q,
N Lu 2
Ǔ
..
and X 4 + * q 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-6)

SPE Drilling & Completion, March 1997 69

You might also like