Sie sind auf Seite 1von 30

Engineering FailureAnalysis, Vol 2, No. 1 pp.

1-30, 1995

Pergamon

Copyright 1995 Elsevier Science Ltd


Printed in Great Britain. All rights reserved
1350-6307/95 $9.50 + 0.00

1350-6307(95)00007-0

STRESS ANALYSIS OF DRILLSTRING THREADED


CONNECTIONS USING THE FINITE ELEMENT METHOD
K. A. M A C D O N A L D and W. F. D E A N S
Department of Engineering, University of Aberdeen, Kings College, Aberdeen AB9 2UE, U.K.

(Received 25 January 1995)


Abstract--Stress analysis of the threaded connections in drillstrings and bottom hole assemblies has contributed to the successful resolution of some downhole failures. The preload
applied from joint make-up torque directly affects the static stress distribution within the
connection--it also affects the local mean stress levels about which the stresses arising from
fatigue loading oscillate. Considering a generic trapezoidal threadform, the relationship
between the nominal applied load and the resulting peak and local elastic stresses at the
critical thread roots are established using the finite element method. The distribution of peak
stress in the connection is determined based on the axial preload arising from make-up, and
how this distribution is modified by tensile and compressive axial loads. Starting with a
procedure of mesh convergence and model validation, a two-dimensional axisymmetric
elastostatic modelling approach is used. In all cases, the roots of the first loaded tooth in the
pin and the last loaded tooth in the box are the sites of maximum peak stress as expected, the
pin peak stress being the greater. However, considering the effects of fatigue loading by
relating the individual preload and tensile load cases to local and peak stress ranges and mean
levels demonstrates that the box becomes the critical component.

1. I N T R O D U C T I O N
Downhole failure of the threaded connections in drillstrings and b o t t o m hole
assembly components by fracture and fatigue, although u n c o m m o n , nevertheless
occurs with sufficient frequency [1] to focus attention on the detail design aspects of
the connections. The primary factors influencing connection failure are the stresses at
critical locations, and the material's fatigue and mechanical properties. Knowledge of
the applied loads during drilling in m a n y circumstances carries with it great uncertainty. Material strength requirements, wellbore geometric constraints, hydraulic flow
area considerations, and economic limitations have all led to the widespread, if not
universal, selection of high-strength low-alloy ( H S L A ) steels for the majority of
drillstring tubular applications [2]. Hence, as the loading regime and material
selection are essentially either predetermined or uncertain, detail design aspects of
threaded connections have a p r e d o m i n a n t role in offering control of fatigue and
fracture performance.
The search for new hydrocarbon resources and the development of existing reserves
are taking place in an economic climate characterised by a depressed oil price and a
concerted industry-wide initiative to reduce costs. Such pressures have dictated
m o d e r n industry trends towards deeper and deviated wells which place m o r e stringent
demands on the design and operation of the load-carrying components of the
drillstring.
Conventional drillstring connections are geometrically complex and, in their
threads, exhibit inherently severe geometric stress concentrations. The criticality with
regard to fatigue that this infers is further exacerbated by the widely acknowledged
differential distribution of the load in such joints. The body of knowledge regarding
load distributions in threads dates from earlier this century [3] and has since been
contributed to by m a n y workers, whose collective efforts have been reviewed
elsewhere [4]. Consideration of the stress concentrations in threads and notches has
also been m a d e [5, 6]. Estimating the fatigue behaviour of a connection--in terms of

K. A. MACDONALD and W. F. DEANS


crack initiation life--is readily accomplished in two stages [7], where the stress
concentration factor (SCF) resulting from the main body stresses and thread bending
stresses is first evaluated, followed by determination of the material's fatigue strength
reduction factor, itself based on the SCF and the notch sensitivity of the material.
Finite element (FE) analysis readily provides the first part of this two-stage fatigue
analysis process by determining the connection's general stress state and critical SCFs.
The objective of the present work is to provide stress analysis data in support of a
wider ongoing research programme concerned with the development of predictive
models of fatigue damage accumulation and limit state failure in drillstring threaded
connections, particularly those in the bottom hole assembly, i.e. drill collars and
stabilisers. Threadforms under consideration include standard API designs (V-0.038R)
in addition to the trapezoidal type reported here.

1.1. Downhole failures


Fatigue has been recognised for some time as an important cause of failure in
general drillstring connections [8, 9], but it is only in recent years that the serious
problems arising from such failures have occurred with sufficient frequency to merit
detailed attention to their cause and prevention [2, 10-14]. Recent North Sea data
from a single operator puts costs associated with downhole separation of the
drillstring at ca 10 million per annum [1]. Investigations of recent failures of
drillstring components [1] have, in the case of the most common casualty connections,
revealed a clear distinction between failure modes and the corresponding failure sites.
Connections (Fig. 1) when failing primarily by fatigue, break consistently in the box
at the root of the last engaged thread (LET) farthest from the seal face (Fig. 2).
However, when the same connection fails entirely by ductile overload shear fracture,
the site of failure is in the pin LET adjacent to the shoulder (Fig. 2). These
observations are based on some 25 connection failures, where, subject to the
difficulties presented by post-separation damage to fracture surfaces and equipment
[1], no material or manufacturing deficiencies were apparent. In section, the morphology of fatigue cracks (Fig. 1) is characterised by an approximate 45 orientation to
the thread flank at positions at and adjacent to the thread root, becoming more
straight (transverse to the drillstring axis) once a distance of more than a few thread
root radii away (Fig. 2). This behaviour is consistent with tooth bending stresses
influencing the short crack, with body stresses becoming dominant at greater crack
depths [15].

1.2. FE analysis of large threaded connections


The severe SCFs in threaded connections give rise to stresses with high peak values
and rapidly increasing gradients approching the thread roots. The accurate estimation
of highly localised stresses in these regions is consequently an exacting computational
task. The task is demanding to the extent that a sufficiently enriched three-dimensional FE model targeted to capture the peak stress and gradient would in practice be
unsolvable with most generally available maniframe computing installations. Novel
solution strategies have been formulated [16] whereby a coarse layer of anisotropic
elements substitutes the threads, replicating the differential load distribution. The
method does not directly yield the peak stresses giving the load distribution only--the
local stresses must be computed subsequently.
The oil and gas industry has widely adopted two-dimensional axisymmetric elastostatic FE analysis in the stress analysis of threaded connections in oilfield tubulars
[17]. This arises primarily from the ability of the method to generate and solve
meshes sufficiently refined to correctly compute the localised thread root stress field,
other benefits accruing from ease of modelling and reduced solution time, although in
some cases it is the only route possible to produce a solution. Elastic analysis is
sufficient unless thread root plasticity is so extensive as to influence the overall load

Stress analysis of drillstring threaded connections

Box connection
vNi

~x~

Pin LET

Box LET

\
\

\
\

Z
!

Pin connection

Drill collar

Rotary shouldered
connection

Fig. 1. Drill collar showing the connection and its failure sites.

distribution by affecting the relative strains within the connection [15], and in any
case is a recommended initial approximation. Axisymmetric models give a parallel
thread representation of the joint and can accommodate axial and radial loading and
boundary conditions, the latter seemingly representing a limitation for simulating
drillstring components in bending. Some FE solver codes now offer axisymmetric
elements capable of asymmetric loading [18] which facilitate non-uniform distributed
loads and allow modelling of applied bending in two-dimensional axisymmetric
models. Inevitably, however, such two-dimensional analyses fail to represent both the
thread helix and the runout regions--important features of threaded connections.
Three-dimensional FE studies of threaded connections have been conducted
[19, 20], investigating the effects of pitch, stress relief features and thread geometry

K. A. M A C D O N A L D and W. F. D E A N S

modifications, with helpful results. However, the models are of necessity generally
coarse and in some cases only a few threads are considered and with simplified thread
profiles.

2. C O N N E C T I O N S T U D I E D

2.1. Geometry
The geometry of the connection studied is shown in Fig. 1. It is a 9 in. outside
diameter drill collar connection but with a generic trapezoidal threadform dimen-

(a)

(b)

Scale

Fig. 2(a) and (b). Caption on p. 5.

cm

Stress analysis of drillstring threaded connections

(c)

Fig. 2. Drillstring service failures: (a) fatigue failure of a 6.625 in. diameter stabiliser pin; (b)
overload failure of a pin from a crossover connection; (c) macrograph showing a fatigue crack
at a thread root.

sioned and pitched to approximately match a standard API NC-61 V-profile threadform (V-0.038R) [21, 22]. An analysis of the NC-61 connection in standard form is
included in this study and may be published at a later date.
2.2. Material
For the purposes of this numerical study, both the pin and box materials were
assumed linear elastic with the general mechanical properties of steel, elastic modulus
E = 207,000Nmm -2 and Poisson's ratio v = 0.29, and for the analyses involving
plasticity, a yield point of 8 0 0 N m m -2 (116 ksi). In reality, the majority of ferritic
drillstring materials are specified according to AISI 4145 H and AISI 4142 H, both
HSLA steels in the quenched and tempered condition. Typical mechanical properties
are given in Table 1.

3. N U M E R I C A L ANALYSIS M E T H O D

3.1. Computational facilities


Preparatory work and the FE analyses proper were performed using A B A Q U S /
Standard, a general-purpose FE program [18], across a number of software releases:

Table 1. Typical specified mechanical properties for AISI 4145 H


Mechanical property

Material specification

Charpy V-notch impact energy


Hardness
Minimum yield strength
Ultimate tensile strength
Reduction of area
Elongation

41 J (30 ft-lbs) at RT
285-341 BHN
758.6 N m m -2 (110 ksi)
965.5 N m m -2 (140 ksi)
45% min
13% min

K. A. MACDONALD and W. F. DEANS


v. 4.9-v. 5.3, Presently, the software is mounted on the Aberdeen University Computing Centre's Sun SparcServer 1000 computer (four 50 MHz CPUs) running the
SOLARIS v. 2.3 operating system, and accessed via SunSparc IPC workstations. The
FE models were prepared using A B A Q U S ' s own model data definitions for the more
simple models, but employed P 3 / P A T R A N [23], a graphics-based mesh-generating
and results post-processing program, for the more refined and complex meshes.
Results post-processing was carried out using both P 3 / P A T R A N and A B A Q U S / P o s t .
Particular features of A B A Q U S exploited in these analyses included geometric
nonlinearity, frictional contact modelling using interface elements, automatic resolution of overclosed interfaces (shrink fitting) and the nonlinear asymmetric deformation capability for axisymmetric solid elements.
3.2. Axisymmetric modelling
The single most important approximation made in constructing the FE model is
that of axisymmetry, where the geometry is considered a body of revolution about the
Z-axis modelled in the R - Z plane (Fig. 1). This approximation significantly
alleviates the computational load which would otherwise be associated with a
three-dimensional analysis and has consequently been widely adopted by the oil and
gas industry [17]. The validity of this assumption is inferred from corroboration
between two-dimensional FE analyses and three-dimensional photoelastic studies
performed on standard (ISO M30) nut and bolt connections [24]. The axisymmetric
representation of the three-dimensional component does not, however, represent
either the thread helix or the runout regions, simply modelling the connection as a
series of parallel threads with a constant runout geometry. A further limitation exists
in that conventional axisymmetric models are limited to axial loads, but bending loads
are a major contributor to the stress state in a drillstring [8]. This apparent restriction
is rationalised by arguments constructed around the remoteness of the connection wall
from the neutral axis (small ratio of wall thickness to diameter) which leads to a
limited stress gradient across the wall under applied bending loads, which consequently can be adequately approximated by a uniform membrane load (Fig. 3). This
assumes no coupling between the regions of maximum tensile and compressive stress,
disposed 180 apart, so that separate analyses of equivalent tension and compression
loads are considered to give reliable results.
A B A Q U S / S t a n d a r d includes in its element library elements suited for the nonlinear
analysis of initially axisymmetric components which undergo non-linear, asymmetric
deformation--as occurs with tooljoints subjected to bending loads. Whereas the
conventional axisymmetric continuum elements use standard isoparametric interpolation with respect to R and Z, such asymmetric-axisymmetric elements have additional Fourier interpolation with respect to 0. The use of such elements allows the
recovery from a single model of stresses in a connection loaded in pure bending, or
bending combined with axial load, and removes the simplification of replacing the
actual non-uniform load with an idealised uniform load. However, the effects of
thread helix and runout are still not considered.

4. MESH C O N V E R G E N C E
The extremes of mesh density can produce an incorrect solution if too coarse, and
analysis costs disproportionate to the results if too fine. A fine mesh is needed in
regions of high stress (and strain) gradient which occur at geometric discontinuities,
where a coarser mesh will suffice in areas of constant stress or low stress gradient.
Furthermore, element formulation is important in that, for a given problem, a linear
displacement element enquires a finer mesh than a parabolic one which, in turn,
needs a finer mesh density than a cubic element.
Because the effect of a stress concentration on the elastic stress field is local and

Stress analysis of drillstring threaded connections

outer fibre stress


stress differen
across wall --,

wall t ickness

inside diameter

outside diameter

. . . . longitudinal
axis

Fig. 3. Through-wall distribution of stress in a pipe under a bending load.

dies away or diffuses with distance, a graduated mesh can be used in such areas [25],
accommodating the transition from a fine mesh at the stress concentration to a coarse
mesh in remote regions. On occasions where the local stress at a particular discontinuity is not of primary interest, but the stress at another site is, a coarse mesh
can be used at the discontinuity and accurate stresses still obtained at the site of
interest (with appropriate local mesh density) provided this site is sufficiently remote
from the discontinuity, in doing so recognising that accurate stresses will not be
obtained in the coarsely modelled region. Such an approach is useful provided the
coarsely modelled region gives correct load paths, stiffness and boundary conditions.
Elements are typically defined in terms of the basic shape of the parent element,
for example a square for a quadrilateral element, an isosceles triangle for a triangular
element and so on. Complex geometries can pose difficulties in controlling element
shape, increasingly distorted elements generally producing less accurate results. In
general, more distortion can be accommodated without loss of accuracy with both
higher-order elements and smaller stress gradients.

4.1. Mesh convergence study


With the aim of refining the mesh at the geometric discontinuity represented by the
thread root, a series of small submodels were constructed covering a range of mesh
densities local to the thread root (Fig. 4). The boundary conditions and loads were
identical for each model: uniform axial pressure (stress) on one component; removal
of the axial (Z-direction) rigid body mode on the other. Radial interference between
the box and pin was also included by setting extreme typical manufacturing tolerances
to give a radial interference of 0.0508 mm (0.002 in.). Due to their reliable performance [18, 24, 25] eight-noded, biquadratic interpolation, reduced integration, axisymmetric quadrilateral elements were used. Contact was modelled using interface
elements between the mating thread surfaces, these elements allowing for closing and
opening of contacting surfaces, small relative sliding and the modelling of friction by
means of the classical Coulomb model [18]. Although a coefficient of friction of 0.09
was selected, being appropriate to drillstring connections [21], this coefficient was
extremed in this study to investigate the sensitivity of the results to this assumption.

K. A. MACDONALD and W. F. DEANS


tJtl

~ - ~ H

mesh

J-

)~

mesh

Fig. 4. FE meshes used for mesh optimisation study.

A l t h o u g h e x t r a c t e d in i s o l a t i o n f r o m the full c o n n e c t i o n , a n d d e s p i t e its simplicity,


the c o n f i g u r a t i o n o f the m o d e l is c o n s i d e r e d a d e q u a t e l y r e p r e s e n t a t i v e o f the real
case, in that the effect of the t h r e a d n o t c h , t o o t h b e n d i n g l o a d s a n d b o d y stresses are
all i n c l u d e d .
T h e a x i s y m m e t r i c F E m o d e l s w e r e s o l v e d for static l o a d c o n d i t i o n s using g e n e r i c
elastic m a t e r i a l p r o p e r t i e s for steel ( E = 2 0 7 , 0 0 0 N m m -2, v = 0.29) by, firstly,
r e s o l v i n g the r a d i a l i n t e r f e r e n c e b e t w e e n the pin t h r e a d a n d b o x , a n d , s e c o n d l y ,
a p p l y i n g a s u b s e q u e n t u n i f o r m axial stress. T h e v a r i a t i o n o f m a x i m u m p r i n c i p a l
stress* a r o u n d the r o o t r a d i u s o f the p i n ' s e n g a g e d t h r e a d is s h o w n in Fig. 5 for a
r a n g e o f m e s h densities, w h e r e it is clear that significant m e s h r e f i n e m e n t is n e e d e d to
a c c u r a t e l y r e p r e s e n t the stress d i s t r i b u t i o n a r o u n d the t h r e a d r o o t radius. This p e a k
stress o c c u r r e d at a b o u t 40 f r o m the t h r e a d f l a n k (Fig. 6). T h e r a d i a l ( t h r o u g h - w a l l )

*From a rigorous theoretical standpoint, the maximum tensile stress at the thread root is located at an
unloaded (free-surface) boundary and hence occurs at the site of the tangential in-plane principal stress at
that point. This theoretical uniaxial stress state may not be recovered exactly by an FIE analysis where the
maximum principal stress is potentially matched with a complementary but spurious non-zero minimum
principal stress normal to the free surface. Taking the maximum principal stress difference in preference to
the absolute maximum principal stress can ameliorate this marginal inaccuracy. However, such errors
amounted to < 0.8% of peak values in the mesh optimisation study: hence, the maximum principal stress
alone was used for convenience.

Stress analysis of drillstring threaded connections


(a)

1600
1

2
1200
cO

16
800

400

-400

0.5

1.5

2.5

Distance,

(b)

~u

mm

1500-

box

&

r'l

pin
i000

<

.,4

500

.M
@
n.

-500
0

,
2

t
3

Distance,

w
4

mm

Fig. 5. Distribution of peak stress around the thread root: (a) mesh optimisation study (pin);
(b) pin and box (mesh 4).

gradients of maximum principal stress approaching the thread roots (Fig. 7) are
consistent for all mesh densities, only diverging close to the thread root where the
stress gradient increases rapidly. Again, the more refined meshes capture the stress
gradient: however, the maximum peak stress is only obtained with the most highly
refined meshes. The accuracy and efficiency of the five meshes studied are clearly
represented in Fig. 8, where the convergence of peak stress is evident, as is the
associated computational cost. The results from a similar series of analyses using

K. A. M A C D O N A L D and W. F. D E A N S

I0

Fig. 6. Contours of m a x i m u m principal stress showing the site of m a x i m u m stress along the
thread root radius.

1600
16

1200

[]

&
-"

.I
800

400

0
0.7

0.75

0.8

Normalised

Fig.

0.85

wall

0.9

0.95

thickness,

mm

7. Stress distributions resulting from the different meshes.

first-order quadrilateral elements are also given in Fig. 8, where the consistent
superiority of the second-order elements is clear. Because the coefficient of friction
used in the analyses ( f = 0.09) is simply selected from standards appropriate to
drillstring connections [21], the effect of a higher value was also studied ( f = 0.5), but
the difference in peak stress proved negligible (0.07%).
Mesh number 4 (Fig. 4), with eight elements defining the root radius, was selected
as the optimised mesh for use in subsequent analyses based on its compromise
between converged peak stress output and reasonable computational cost.

Stress analysisof drillstringthreadedconnections

11

CAXSR stress

13

---13---

CAXSR time

CAX4R stress

&

-'-&-'-

CAX4R time

8OO

1600

S
I Q

14oo

700

-r'l

1200

600

lOOO

5OO

800 ~

"//s/"

--~

O
-r'l
r"4
O
r.t)

4OO
'300

600

400

,.o,A

-200
-I00

200
0

~.ii~ ~
--

..41~.~.
I

i0

15

Number

of

0
2O

elements

Fig. 8. Mesh optimisation results: dependence of peak stress and solution time on mesh
refinement.

5. AXISYMMETRIC MODEL VALIDATION


5.1. Thread runout geometry

Representation of a three-dimensional threaded connection by a two-dimensional


axisymmetric FE model inevitably fails to include the effects of thread helix and
runout geometries. These geometric features combine in the critical regions of a
connection--the pin and box LETs--to produce a nonuniform thread load distribution around the runout thread progressing from the first contact when partly formed
to a fully developed inter-tooth load as the helix progresses and the thread chamfer
diminishes, typically occurring in one complete revolution. The increased flexibility of
the first partly formed thread over its fully formed neighbours initially encourages
load shedding from the mating thread, until the thread becomes completely formed,
attaining the peak thread root stress about one turn from the point of initial
inter-tooth contact [26]. An attempt to model such behaviour in axisymmetry can be
made by considering the thread runout geometry at several meridional sections,
equi-spaced at angular sites. Although it is impossible to argue that such an approach
embodies the actual stress state, the method has nevertheless been validated with
experimental data for standard nut-bolt connections [24, 26], and as such is
considered a viable approximation for use in the present study.
A typical FE mesh, from a series of three, representing different meridional planes
through the whole connection, is shown in Fig. 9. The level of mesh refinement is
identical at all threads--as the peak stress distribution for all threads is of interest-and is based in the optimised mesh considered earlier. The typical model size was
approximately 13,600 elements and 40,000 modes. These models were solved for
resolved radial interference and axial tension loading, with the Z-direction rigid body

K. A. M A C D O N A L D and W. F. DEANS

12

R[

Fig. 9. FE model of a full connection (perimeter plot) with pin LET: details of uniform and
differentially meshed models.

13

Stress analysis of drillstring threaded connections

mode restrained on the pin free end. Substructuring with superelement generation
was not employed because ultimately the modes were intended to be solved for
non-linear material behaviour and non-linear asymmetric-axisymmetric deformation.
The differences in thread runout geometry gave rise to variations in peak stress (Fig.
10). In contrast to results from studies of standard V-form threads [24] the
multi-meridional plane technique produces the maximum stress in the thread at a
point close to the first attainment of full thread height on the contacting flank, i.e.
where the thread is still not yet fully formed (Fig. 9). The peak stress occurs at this
location due to the superposition of the maximum connector body stress and the

(a)

1250

I000

.~

750

500

250

0
0

Tooth

(b)

pitches

from

shoulder

1250 -

d
i000 -

750 -

500 - A

-_.:$

. . . . . .

25O

i0

Ii

12

13

14

15

Tooth

Fig. 10.

pitches

from

seal-face

Thread peak stress distribution from three separate multi-meridional plane FE models
with different thread runout geometries: (a) pin; (b) box.

14

K. A. M A C D O N A L D and W. F. D E A N S

maximum possible tooth load bending stress. Furthermore, the maximum pin and box
stresses do not necessarily occur in the same meridional plane because of non-identical thread runout formation in a given model. Based on these results, a single
axisymmetric geometry was chosen with similar states of runout evolution on both the
pin and box, maximising the thread peak stresses in both.

5.2. Preload from make-up


Established drilling practice is to further tighten connections after assembly by
applying a make-up torque. In addition to simply removing slack in a joint and
providing resistance to downhole make-up arising from shock torsional loading and
the reduced coefficient of friction of pipe dope at elevated downhole temperatures,
this practice results in strong benefits from a live load carrying perspective. Once
preloaded, axial tension is generated in the pin which is equilibrated by axial
compression in the box. Seemingly, an alternative load path is created through the
shoulder interface, bypassing the pin threads [15]: however, in reality applied loads
are still transferred through the pin, causing it to stretch or shorten and the shoulder
compression to release or increase. The complexities of modelling the relationship
between applied torque and axial preload is sufficiently difficult that, in this study,
use was made of prescribed axial preloads from drillstring standards. The recommended make-up torques used for connections are taken from drill collar
guidance [21], where specified torques are intended to generate a minimum axial
stress of 62,500 psi (431 N mm -2) in the weaker of the pin or box.
In FE modelling terms, overclosed interface elements were used at the seal-faceshoulder region which, when resolved as an interface fit, produce a compressive
interfacial pressure which reacts in the threads as an axial preload. Furthermore,
interface elements allow the shoulder interface pressure to vary in accordance with
applied loads, mimicking the make-up preload mechanism in practice.
A series of identical FE models of the full connection with increasing amounts of
shoulder overclosure were solved by resolving the overclosure. An overclosure of
0.3 mm p r o d u c e d - - f r o m a linearisation procedure -- a membrane stress component of
467 N m m -2 located in the parallel section of the pin adjacent to the shoulder (Fig.
11), and was thus considered to compare well with the API guidance. This value of
0 . 3 m m seal-face overclosure compares favourably with a value of 0.008in.
(0.203 mm) used elsewhere for an analysis of a standard NC-46 connection [27]. The
resultant peak stress distribution (Fig. 12), demonstrates that the pin L E T and the
other lower thread numbers* react with most of the axial preload.

5.3. Differential mesh density at thread roots


Building on the earlier explanation that coarse modelling of stress concentrations
may not prevent a refined mesh elsewhere returning accurate results at the stress
concentration of interest, a second mesh was prepared based on the same axisymetric
geometry as in Sections 5.1 and 5.2 but, unlike its predecessor, it was targeted to give
accurate results at the critical threads only (Fig. 9), with a more coarse mesh at the
intermediate threads. This was done to provide an FE model of reduced size and
attendant reduced computational load for use in analyses where only the critical
threads will be of interest, as is the case with comparative studies. The original model
size was dramatically reduced to about 4000 elements and 11,500 nodes. In solving
the model for combined preload and axial load, comparison with the corresponding
uniformly meshed model shows that the peak stress results at the intermediate threads

*Thread numbers counted from the shoulder.

Stress analysis of drillstring threaded connections


1250

i000

actual
distribution

@
o

15

750

linearisation
500
,<
250

0.2

0.4

0.6

0.8

Normalised

wall

thickness

Fig. 11. Preload induced through-wall distribution of stress at the pin parallel section (0.3 mm
shoulder-seal-face overclosure).

5000 -

box

4000 -

-~I

pin

3ooo-

i 2000-.
m

i000 -

OI

I0

12

14

Tooth

pitches

from

T7
16

shoulder/seal-face

Fig. 12. Thread peak stress distribution solved for preload (0.3 mm seal face overclosure)
from three runout geometries.

are significantly in e r r o r (34 and 18% u n d e r p r e d i c t i o n for pin and box, respectively),
but, m o r e importantly, the results at the critical threads are in g o o d a g r e e m e n t
(within 0.41%). These results validate the n o n - u n i f o r m l y m e s h e d m o d e l , d e m o n s t r a t ing that the coarsely m o d e l l e d region gives correct load paths, stiffness and b o u n d a r y
conditions.

16

K.A. MACDONALD and W. F. DEANS

5.4. Non-linear asymmetric-axisymmetric elements


The fundamental premise of the axisymmetric models dicussed above is that
drillstring loads can be adequately represented by uniform axial loads at level
equivalent to the average membrane stress component across the pipe wall, or the
extreme outer fibre stress (Fig. 3). This approach requires separate analyses of equal
but opposite sign (tensile and compressive) axial loads to r~cover the peak stress
distributions disposed at the 0 and 180 positions. Elements exist which allow the
non-linear analysis of initially axisymmetric components which undergo non-linear,
asymmetric deformation. In essence, such elements allow in-plane bending loads,
assumed to be symmetrical about 0 = 0 , to be applied to axisymmetric models,
overcoming the need for simplification of the stress distribution and for separate
solutions. Results can be recovered at various angular positions depending on the
number of Fourier modes employed, the minimum being two nodal planes at 0 and
180, and the maximum being five nodal planes at 0, 45, 90, 135 and 180 . Bending
loads can be combined with uniform axial loads to represent downhole loads
comprised of a weight-on-bit or overpull component, and a bending component from
a dog-leg or deviated wellbore. As the analysis is non-linear, a superelement solution
strategy cannot be adopted. The analysis is still in two dimensions and the effects of
thread helix and runout are still ignored.
To reduce the computational load, the differentially meshed model was employed
with asymmetric-axisymmetric elements and solved for preload and an equivalent
bending load (maximum outer fibre stress = uniform membrane stress from the
axisymmetric analysis). Comparison with the corresponding results from the axisymmetric solution revealed significant differences in peak stress, particularly at the pin
L E T under preload and the box L E T under preload plus bending. Comparison of the
through-wall stress gradients at the straight section near to the pin shoulder showed
that the asymmetric-axisymmetric elements returned lower stresses at near-surface
positions than the plain asymmetric elements, although the general distributions were
similar (Fig. 13). The same is true at the pin L E T site, where the discrepancies at the
extremities are amplified in the region of high stress gradient near to the thread root.
134
CAX8R
r~
~4
rO

1.4

c6

Normalised

wall

thlckness

Fig. 13. Differences in through-wall distribution of stress at the pin parallel section returned
by plain axisymmetric (CAX8R) and asymmetric-axisymmetric (CAXA8R1) element types
under preload only.

Stress analysisof driUstringthreaded connections

17

5.5. Elastic-perfectly plastic material behaviour


In order to test the contention that an elastic analysis is sufficient unless thread root
plasticity is so extensive as to influence the overall load distribution within the
connection (by affecting the relative strains between threads), comparative analyses
were performed with both elastic material behaviour, and with elastic-perfectly
plastic behaviour using the Von-Mises yield surface. The elastic-plastic material
model used an identical elastic modulus value of 207,000 N mm -2 and, beyond first
yield at 800 N mm -2, assumed a constant value of yield stress (perfectly plasticity).
Analysing a case of preload and 200 N mm -2 applied tension with the uniformly
meshed model, the extent of plasticity at the pin LET was found to be small
[~0.261 mm (Fig. 14)] with the through-wall stress gradients showing close agreement, only diverging once yielding is promoted in the high-stress region local to the
thread root. The extent of plasticity at the next pin thread was smaller still,
0.070 mm, with all other pin and box thread roots remaining elastic. The non-uniform
mesh was also analysed for elastic and elastic-plastic material behaviour with the
results demonstrating excellent agreement with the uniform cases (Fig. 15). In
comparing the merits of the two material models, it is clear that plasticity is
sufficiently limited so as not to contribute significantly to the relative strains within
the joint, and as the analysis costs of the elastic-plastic model are a factor of 5
greater than that of its elastic counterpart, it is concluded that the assumption of
elastic material behaviour is justified in this case.

6. RESULTS AND DISCUSSION


The present interest lies in evaluating the stress distribution and stress concentration factors for a drillstring connection with a trapezoidal threadform under preload
and preload plus uniform axial loads. Concerns over the discrepancies found between
the preload cases modelled with axisymmetric and asymmetric-axisymmetric elements
led to their suspension from this study, and has prompted closer examination of this

lz

Fig. 14. Extent of plasticity at the pin LET for preload plus 200 N mm-2 axial tension
(0.261 mm).

K. A. M A C D O N A L D and W. F. DEANS

18

m~

1 rq(lO

134

--

d
120C)
m
0

X
A

Elastic
Both m e s h e s

+
$
/~"

Elastic-plastic
Differential mesh
--

.Ilk.
/ ' ~

Elastic-plastic

E
o
o

q(}O

O]

~J

c~O0

,--I
.,--4

c<

~ 0{}

() . .~

{) . .1

Hormalised

I
/

. (~

wall

() . ,q

thickness

Fig. 15. Differences in through-wall distribution ot stress at the pin LET returned by plain
axisymmetric (CAX8R) and asymmetric-axisymmetric (CAXA8R1) element types in uniform
and differentially meshed models solved for preload plus 200 N mm 2 axial tension.

element type and its implementation. The results discussed here for the plain
axisymmetric FE analyses are in the most part modelling make-up preload combined
with uniform axial tension and compression loads.
In assessing the effect of a logical modification to the threadform, a comparative
study considered the connection with matched thread root radii on both pin and box
threads, and a modified thread with the root radii increased. These uniformly meshed
models were analysed for axial tension without preload and the results clearly
demonstrate the reduced thread root peak stresses in all the box threads (Fig. 16).
The main body of analyses considered the uniformly meshed model under preload
(0.3 mm shoulder overclosure) combined with uniform axial tension and compression
at nominal stress levels ranging from + 100 to + 800 N m m - : , representing a range
approximately 12.5-100% of assumed material yield strength. The primary results in
terms of the distribution of thread root peak stress demonstrate that the preload
effects dominate the overall distribution. The pin LET consistently shows the highest
peak stress and through-wall stress gradient (Fig. 17). On first inspection the results
for the box are unusual in that the first engaged thread (FET) exhibits a higher peak
stress than the LET--a result apparently contradicting service experience. However,
examination of the stress gradients at the box FET and LET positions (Fig. 18)
reveals that, although the FET peak stress is indeed high,* the through-wall gradient
of stress is largely compressive, only becoming slightly tensile once axial loads reach
levels which promote shoulder separation (Fig. 19). In contrast, the box LET with a
lower peak stress exhibits a consistently tensile through-wall gradient, confirming it as
the known site of failure from service experience.
A vector representation of the thread root stress field demonstrates the variation of
maximum principal stress direction (Fig. 20). The maximum principal stress direction
is initially tangential to the root radius at the point of peak stress, but rotates to
become axially aligned once more than a distance equivalent to a few root radii away
*The existence of this highly localised SCF at the box FET has been confirmed by metallographic
examination, where loealised damage to the thread root has been observed.

Stress analysis of drillstring threaded connections

19

800-

E
~

600

200

I
6

I
8

I
10

I
12

I
14

I
16

I
18

T o o t h pitches from seal-face


Fig. 16. Effect of enlarging the box thread root radius on the thread peak stress distribution
(axial tension Nmm-~; no preload).

from the thread root. This gives rise to an initially curved crack propagation path
near to the thread root which produces a characteristic lip feature on the failure
surface. This fatigue crack morphology potentially provides an important point of
reference for failure investigations as would be readily observed even when postfailure mechanical damage has taken place [1]. These results exemplify the competing
mechanisms of tooth bending stress and the stress concentrating effect of the tooth
notch on the body stresses [5, 15].
Tooth separation occurs at the higher tooth numbers under compressive loading,
demonstrated by the tooth flank interface element openings (Fig. 21), where the
spread of tooth disengagement from the box LET towards the pin LET at increasing
compressive loads is clear. Radial interface displacements between the pin and box
threads under preload and tension loading demonstrate the extent of radial expansion
and contraction of the connection* (Fig. 21), the greatest changes from the preload
state taking place at the connection extremities due to the high proportion of load
transfer and increased flexibility arising from the taper at these sites.
6.1. SCFs
In preloaded connections, the representation of the stress state local to the critical
thread root using classical SCFs is substantially dependent on how the SCF is defined.
Evidenced by Fig. 12, the preload has a pronounced effect upon the stresses at low
thread numbers while the preload remains in force. The notch local stress, defined as
the sum of linearised bending and membrane through-wall stress components, displays
non-proportional behaviour in the case of the pin where the LET local stress is clearly
a nonlinear function of nominal pipe stress (Figs 22 and 23). The box LET local stress
remains largely unaffected by the preload and consequently exhibits a proportional
response to applied stress. The change in gradient of the pin local stress response at
*This relative radial displacement is measured between the crest of the box thread and the root of the pin
thread, and is composed of both increases and decreases in radial displacement of the box and pin,
respectively.

20

K.A. MACDONALDand W. F. DEANS

[]

600 MPa
400 MPa
200 MPa
100 MPA
preload

i1200;MM~:

-i

-400 MPa
-600 MPa

-at
(a)

7000 -

6000-

aa
5000

-,-I

4000 -

3000

<

2000

i000 -

Tooth

~
-

t~

4000 1

pitches

~.
-at,
[]

600 MPa
400 MPa
200 MPa
100 MPa
preload

-100 MPa
-200 MPa
-400 MPa
-600MPa

at

4%

~ ~

3000]

1 ~)

! 2

14

from

I
16

shoulder

2000

1i]
II

Tooth

i0

pitches

12

from

14

seal-face

Fig. 17. Effectof axial load oll the thread peak stress distribution:(a) pin: (b) box.

16

Stress analysis of drillstring threaded connections


n.

I000

800

600

400

21

[]

FET

LET

200
4J
m

r6

.,-I

-200
-400
-600
0

0.2

0.4

0.6

0.8

Normalised

wall

thickness

Fig. 18. Differences in through-wall distribution of stress at the box FET and LET sites
(preload plus 200 N mm -2 axial tension).

tension
1250

[]

compression

d
i000

75O
0

500
I--4

250

200

400

600

Axial

stress,

800

MPa

Fig. 19. Effect of applied axial tension and compression on shoulder-seal-face interface
pressure.
n o m i n a l a p p l i e d stresses o f a p p r o x i m a t e l y
p o i n t s w h e r e t h e t o o t h p r e l o a d is o v e r c o m e
is r e l e a s e d in t e n s i o n . T h e v a r i a t i o n of
b e h a v i o u r in t h a t t h e b o x L E T r e s p o n s e

___450 N m m -2 (Fig. 22) d e f i n e s well t h e


in c o m p r e s s i o n a n d t h e s h o u l d e r p r e l o a d
n o t c h p e a k stress d e m o n s t r a t e s s i m i l a r
is effectively linear* b u t the pin L E T

22

K. A. M A C D O N A L D and W. F. D E A N S

{
I

\\\

"" \

'

\
3

Fig. 20. Variation of m a x i m u m principal stress direction within the complex thread root stress
field (preload plus 200 N m m - : axial tension).

response is evidently nonlinear (Fig. 22). Thus, the peak stresses cannot be expressed
as constant geometric SCFs. In order to describe these peak stresses with respect to
some reference stress within the connection, three appropriate definitions of SCF (Kt)
can be used:
K t relative to the nominal pipe membrane stress. Although the most conveniently derived SCF, this classical geometric definition does not accommodate
the potentially non-proportional relationship between the peak and nominal pipe
stresses (Fig. 22) and fails to take account of the preload's profound effect upon
the SCF.
(ii) Kt relative to the local membrane component of stress. Linearisation of the
through-wall stress distribution at the critical thread (Fig. 23) to give the
membrane stress component takes account of the body stresses and gives a better
representation of the SCF.
(iii) Kt relative to the local membrane and bending components of stress. Although
more onerous in terms of data reduction, a full linearisation of the through-wall
stress distribution at the critical thread (Fig. 23) evaluates the thread root peak
stress relative to the most effective measure of the notch nominal local stress.
(i)

Considering first the box LET, Fig. 24 gives the SCFs computed on the basis of the
three above definitions as functions of nominal applied pipe stress. The SCF based on
pipe stress is a weak function of applied stress, but when expressed with reference to
the sum of local membrane and bending stress components it becomes effectively
constant at a value of approximately 3.5 in tension, even when the preload mechanism is substantially overcome at pipe stresses above about 450 N mm -2. In compression, the box L E T is no longer loaded by a meshing pin tooth but is simply subjected
to compressive body stresses: consequently, the stress state alters accordingly and a
lower SCF of 2.0 is the result.
*Under compressive load, the box L E T disengages and is no longer loaded. In this condition, the stress
state is characterised by the box body in compression and the site of compressive peak stress moves from
the former site of the tensile peak stress to another position on the thread root.

Stress analysis of drillstring threaded connections

(a)

0.35

[]
A
--

0.3-

23

100 MPa
200 MPa
400 MPa

0.25 -

,,~

0.2-

0.15 Q)

~Z

0.i-

0.05 -

I
0

pitches

Tooth

0.14

(b)

from

i0

12

14

16

shoulder/seal-face

o.12-

0.i-

preload
100 MPa
200 MPa
400 MPa

t
l/
[
I
]
/

0.08 J.a
in

@
o

0
0
b
~Z
0
0

0.06 -

0.04 -

0.02 -

0
0

Tooth

pitches

from

i0

12

14

16

shoulder/seal-face

Fig. 21. Effect of load on the thread flank and crest interface openings: (a) flank; (b) crest.

Identical but magnified effects are apparent for the pin LET in tension (Fig. 24),
which might be expected as the preload mechanism has been shown to have most
effect at this site. The SCF in tension with reference to pipe stress is profoundly
affected by applied stress while the preload remains effective at applied stress levels
below about 450 N mm -2. Computed on the basis of the linearised local bending plus
membrane stresses, the pin SCF reduces to a linear and very weakly dependent
function of applied stress, almost constant at a value of approximately 5.5. Similar
behaviour persists into the compressive loading regime but, unlike the box LET, the
pin LET remains initially engaged by the preload mechanism, with the pin SCF
reducing once the preload is fully overcome at applied stresses below - 4 5 0 N mm -2.

K. A . M A C D O N A L D

24

(a)

i g 08 n

3.

a n d W . F. D E A N S

l
r~

~000

o~

2500

c~

2 :. 0 0
~':~:

I
-60:;

I
20:i

I
200

Nominal

1
o~U-P'-

applied

I
90@

stress,

MPa

(b)
,~

2000

pin

11.

box

1500

i000

500 -

O
0

500 -

-i000

1500
-900 .

. -60,-: .

. ~i,~

I
.

.-

Nominal

2 0 0.

64r,

:,~,,

applied

stress,

MPa

Fig. 22. Effect of applied load on LET: (a) peak: (b) l o c a l stresses.

Once the dominant effects of preload are adequately considered, the SCFs for
tension loading become effectively constant at 3.5 for the box LET and 5.5 for the
pin LET (Table 2), reflecting the greater thread root radius of the box postulated in
this study.
6.2.

Response of local stress to cyclic loading

Cyclic fluctuation of the applied pipe stress promotes an oscillatory response in the
local stresses at the critical threads. These alternating through-wall stresses at the pin

Stress analysis of drillstring threaded connections


--

--.

25

peak stress

- -

Pb

Pm+Pb
linearised stress distribution

Stress distribution at

Pm

At

AA

Fig. 23. Stress linearisation procedure to give stresses used in the definition of SCFs.
and box L E T sites are the driving force for fatigue crack initiation and propagation
behaviour, and are the foundations on which the notch peak stresses are based.
Considering the axial tension load F E analyses, and the preload case, stress ranges
and mean levels were computed for the local stresses at a stress ratio* of R = 0. Of
particular note are the pin L E T responses in range and mean where clearly
observable changes in slope take place at approximately 450 N mm -2 nominal pipe
stress (Fig. 25). As with the peak stresses, this behaviour represents the point at
which the preload is overcome and the seal-face-shoulder interface begins to open.
The box L E T has consistently higher local stress ranges than the pin L E T across
the full extent of the applied stress ranges (Fig. 25). The difference between the two
initially diverges but then converges with increasing applied stress range. Such
behaviour occurs because the pin L E T receives most of the protection afforded by the
preload mechanism [15, 28] whereas the box L E T receives virtually none and as such
is exposed to the full effect of the applied stress range. The benefit afforded the pin is
lost once the preload is overcome, accounting for the convergence of the pin and box
responses at higher stress ranges. Notably, the limited stress range at the pin L E T is
apparently obtained at the considerable expense of a consistently elevated mean level
(Fig. 25) much higher than the mean level experienced by the box LET.
6.3. Response of peak stress to cyclic loading
The peak stresses are in fact idealised elastic stresses and as such these extremely
high magnitudes do not occur in practice due to localised yielding. The effect of the
stress concentration is to form a localised region of plastically deformed material
which is best characterised by a strain parameter and not stress. In any case, it is
alternating plastic strain that is acknowledged as the mechanism promoting the fatigue
crack initiation process [29]. These points combine to make a peak stress representation of cyclic loading response unsuitable. However, for the straightforward case of
pulsating tension at R = 0, it is nonetheless both justifiable and helpful to consider
the range and mean levels of peak stress in order to examine the relative severity of
the pin and box L E T sites under fatigue loading (Fig. 26). As with the local stresses,
*Stress ratio definitions: R = -1 is fully reversed tension-compression; R = 0 is pulsating tension from
zero; R > 0 (positive R ratio) is tension-tension.

26

K.A. MACDONALD and W. F, DEANS


(a)

<)
Oq

i
-"
a

~.__

1
-900

O
g]

-]
:3

;. . . .

-600

pipe stress

Pm
Pm+Pb

"

300

'~;

Nominal

applied

(b)

300

600

pipe

stress,

MPa

pipe stress

Pm

Pm + Pb

'[

-i

Nominal a p p l i e d
Fig. 24. Effect of applied

900

pipe

stress,

MPa

load on connection stress concentration factors: (a) box: (b) pin.

the box LET exhibits the highest stress range whereas the pin LET presents the
highest mean level. As demonstrated earlier, plasticity is promoted at the pin LET by
the preload (Fig. 14), and the superposition of further alternating load results in cyclic
plastic strain about a high mean level at this site. Although operating about a lower
mean level, the box LET will undoubtedly also be subject to local plastic deformation
at larger stress ranges but with a greater range of plastic strain.
The in-service fatigue damage already referred to is by definition the whole fatigue
life (initiation plus propagation), which means that, although meaningful conclusions
can be drawn with regard to fatigue initiation life from the peak stress numerical

Stress analysis of drillstring threaded connections

27

Table 2. SCFs in tension for a preloaded connection in a 9 in. diameter drill collar with a trapezoidal
threadform
Nominal pipe stress
(Nmm -~)

Pin SCF

100
200
300
400
500
600
700
800

Box SCF

Pipe

Pm + Pb

Pipe*

Pm + Pb-t

50.7
27.1
18.9
14.6
12.2
11.1
11.1
11.0

6.5
6.1
5.8
5.5
5.5
5.1
4.9
4.9

7.0
6.3
6.0
5.9
5.8
5.8
5.7
5.8

3.7
3.5
3.5
3.4
3.4
3.4
3.4
3.4

*Constant value assumed = 5.5.


?Constant value assumed = 3.5.
results p r e s e n t e d here, the c o r r o b o r a t i n g service data refers to whole fatigue life and,
as such, a c o m b i n e d and qualitative interpretation of the p e a k and local stress results
is justified. In this respect, of the two critical threads, the box L E T displays the
higher local and p e a k stress range but the lower m e a n level. T h e relative impact on
c o n n e c t i o n fatigue is then primarily a function o f the ferritic material's response to
m e a n stress effects. F o r similar structural steels the stress range tends to be m u c h
m o r e significant than the m e a n level [30, 31]. Fatigue loading then induces a critical
failure site at the box L E T u n d e r fatigue loading at R = 0, but, u n d e r static tension
loading, the pin L E T has the greatest local and p e a k stress levels m a k i n g it the
anticipated critical failure site w h e n the joint is statically l o a d e d (summarised in
Table 3).

7. C O N C L U S I O N S
T h e stress distributions and stress c o n c e n t r a t i o n factors in a drillstring t h r e a d e d
c o n n e c t i o n with a trapezoidal t h r e a d f o r m have b e e n evaluated using the F E m e t h o d
(a)
1400

1200

d
i000

800
600

4OO

2OO

200

Nominal

400

applied

Fig. 25(a). Caption on p. 28.

600
stress

800
range,

MPa

[]

pin

&

box

28

K. A. MACDONALD and W. F. DEANS

(b)

nJ

i400
pin

1200

,I.

i000

0)

800

box

600

400

2 O0 -

. ,I

f!, (3

Nominal applied stress range,

MPa

Fig. 25. Effect of cyclic loading R = () on the LET local stress: (a) stress range; (b) mean
level.
supported by a rigorous model validation exercise. The non-uniform distribution of
idealised elastic peak stress was in agreement with existing analytical, experimental
and numerical data for generic threaded connections. The classical stress concentration factor was found to be inconstant and a decreasing function of nominal applied
load in tension. Allowing for the non-linear relationship between applied stress and
tooth notch peak stress, and for the effects of make-up preload, constant stress
concentration factors of 3.5 and 5.5 were derived for the box and pin L E T positions,

(a)

7000 ]
d

6000

[]

pin

,L

box

C
5000 1
m

4000

ooo1

.//

i001~0

200

400

600

800

Nominal applied stress range, MPa


Fig. 26(a). Caption on p. 29.

Stress analysis of drillstring threaded connections


(b)

29

7000
6000

pin

box

5000

4000

3000

2000

i000

200

400

600

800

Nominal

applied

stress

range,

MPa

Fig. 26. Effect of cyclic loading at R = 0 on the LET peak stress: (a) stress range; (b) mean
level.
Table 3. Stresses at critical threads under static and cyclic loading
Cyclic pipe stress
range (0-400 N mm -2)
Static pipe stress
(400 N mm -2)

Peak stress

Location

Peak stress

Local stress

Pin LET
Box LET

I 5857
2350

1073
686

Local stress*

Range

Mean

Range

Mean

1156

5279
1244

388
~

879
351

I
~

*The sum of linearised membrane and bending stress components.

respectively, reflecting in part the greater t h r e a d r o o t radius of the box i m p l e m e n t e d


in this study. T h e stress state at the t h r e a d r o o t is characterised by a varying
m a x i m u m principal stress direction which p r o d u c e s a characteristic lip feature on the
failure surface, potentially providing an i m p o r t a n t point o f reference for failure
investigations.
U n d e r cyclic loading in pulsating tension (R = 0), the p r e l o a d e d joint analysed
p r o d u c e d local and p e a k stress ranges m a r k e d l y r e d u c e d at the critical pin thread, but
oscillating a b o u t high m e a n levels. In contrast, the critical box t h r e a d exhibited larger
local and p e a k stress ranges but with lower m e a n levels. Overall fatigue p e r f o r m a n c e
of a p r e l o a d e d c o n n e c t i o n is thus a c o m p r o m i s e b e t w e e n decreased stress ranges and
increased m e a n levels.
P r e l o a d directly affects the static stress distribution within the c o n n e c t i o n m a k i n g
the pin L E T the critical failure site for static loading, but it also affects the local and
p e a k m e a n stress levels and stress ranges arising f r o m fatigue loading, m a k i n g the box
L E T the expected critical site for fatigue loading at R = 0.

Acknowledgements--Particular thanks are due to D. M. R. Bell of the Aberdeen University Computing


Centre for his expertise and assistance with the University's computing facilities, and to I. Mackinnon of the
Department of Engineering for reprographic services.

30

K . A . MACDONALD and W. F. DEANS

REFERENCES
L. K. A. Macdonald, Engng Failure Analysis 1, 91-i 17 (1994).
2. M. B. Kermani, Proceedings of the International Conference on Environment Assisted Fatigue, Sheffield
(1988).
3. M. B. Kermani, Fatigue of Large Diameter Threaded Connections (edited by W. D. Dover, P. J.
Haagensen, S. Dharmavasan and G. Glinka), Howard Lee, London (1988).
4. E. A. Patterson and B. Kenny, Fatigue of Large Diameter Threaded Connections (edited by W. D.
Dover, P. Haagensen, S. Dharmavasan and G. Glinka), Howard Lee, London (1988).
5. R. B. Heywood, Proc. IMechE 193,384-391 (1948).
6. H. Neuber, Kerbspannungslehr (2nd edn), pp. 159-163, Springer, Berlin (1958).
7. E. A. Patterson, Fatigue Fractures Engng Mater. Struct. 13, 59-81 (1990).
8. A. Lubinski, JPT February, 175-194 (1961).
9. H. M. Rollins, AAODC Rotary Drilling Conference, Dallas, TX (21 February 1966).
10. M. C. Moyer and B. A. Dale, JPT May, 982-986 (1984).
11. S. D. Hampton, Proceedings qf the SPE/IADC Drilling Conference, New Orleans, LA. pp. 177-189,
SPE 16072 (1987).
12. B. A. Dale, SPE Drilling Engng December, 356-362 (1988).
13. C. E. Stromeyer, lnst NavalArchit. Trans. No. 60, 112-122 (1918).
14. B. A. Dale, SPE Drilling Engng September, 215-222 (1989).
15. D. A. Topp, Fatigue of Large Diameter Threaded Connections (edited by W. D. Dover, P. Haagensen,
S. Dharmavasan and G. Glinka), pp 1-15, Howard Lee, London (1988).
16. J. L. Bretl and R. D. Cook, Int. J. Numer. Meth. Engng 14, 1359-1377 (1979).
17. W. D. Dover, P. Haagensen, S. Dharmavasan and G. Glinka, Fatigue of Large Diameter Threaded
Connections, Howard Lee, London (1988).
18. Hibbitt, Karlson and Sorensen, Inc. ABAQUS User's, Examples, Theory and Verification manuals,
HKS, Providence, RI (1992).
19. H. C. Rhee, Proceedings of the OMAE 90, Houston, TX, 18-23 February, pp. 293-297 (1990).
20. A. Tafreshi and W. D. Dover, Int. J. Fatigue 15,429-438 (1993).
21. API Recommended Practice 7G (RP 7G) Recommended Practice for Drill Stem/Design and Operating
Limits" (14th edn), American Petroleum Institute, Washington, DC (1990).
22. API Specification 7 (Spec 7) SpeciFication [br Rotary Drill Stem Elements (38th edn), American
Petroleum Institute, Washington, DC (1994).
23. PDA Engineering, P 3 / P A T R A N Release Notes, Release 1.3-2. Costa Mesa, CA (1994).
24. E. Dragoni, J. Offshore Mech. Arctic Engng 116, 21-27 (1994).
25. NAFEMS, A Finite Element Primer (2rid reprint) (1991).
26. E. A. Patterson and B. Kenny, J. Strain Analysis 21, 17-23 (1987).
27. J. E. Smith, Fatigue of Large Diameter Threaded Connections (edited by W. D. Dover, P. Haagensen,
S. Dharmavasan and G. Glinka), pp. 161-188, Howard Lee, London (1988).
28. A. Newport and G. Glinka, J. Engng Mech. 17. 1257-1273 (1991).
29. J. F. Knott, Fundamentals of Fracture Mechanics, Butterworths, London (1981).
30. S. J. Maddox, Fatigue Strength of Welded Structures (2nd edn), Abington Publishing, Cambridge
(1991).
31. T. R. Gurney, Fatigue of Welded Structures (2nd edn), Cambridge University Press, Cambridge (1979).

Das könnte Ihnen auch gefallen