Engineering Failure Analysis 74 (2017) 11–34

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

Fatigue crack growth analysis of drill pipes during rotary drilling

operations by the multiple reference state weight function
approach
Choja Ojanomare a, Pietro Cornetti b, Raffaele Romagnoli a, Cecilia Surace b
a
Politecnico di Torino, Department of Environment, Land and Infrastructure Engineering, corso Duca degli Abruzzi 24, 10129 Torino, Italy
b
Politecnico di Torino, Department of Structural, Building and Geotechnical Engineering, corso Duca degli Abruzzi 24, 10129 Torino, Italy

a r t i c l e i n f o a b s t r a c t

Article history: Drill-string fatigue failures are the most common and costly type of failures in Oil & Gas and
Received 6 May 2016 Geothermal drilling operations. However, the drilling industry currently does not have a
Received in revised form 28 December 2016 means of quantitatively evaluating drill-pipe fatigue fife. The aim of the present study is to
Accepted 29 December 2016
fill this gap, providing a rapid and easy means of accurately obtaining drill-pipe fatigue life es-
Available online 30 December 2016
timates. The geometry considered is the passage of a drill-pipe through a dogleg. It is
characterised by a load combination of tension and bending while undergoing rotation. This
Abbreviations: load–geometry combination is taken as representative of the present-day drilling. The Stress
FCG
Intensity Factors along the crack front of a semi-elliptical internal crack in a drill-pipe are ob-
Topic:
tained by means of Multiple Reference States Weight Function approach that is then imple-
Fatigue Crack Growth
MRS mented in a developed MATLAB code. The developed computer code is validated by
Topic: comparison with data available in the literature and then applied to the geometry of interest.
Multiple Reference States The Stress Intensity Factors thus obtained are then inserted in the Paris Fatigue Crack Growth
SIF Model to quantitatively evaluate the drill-pipe fatigue life. Finally, a parametric study is per-
Topic: formed in order to identify the influence of load magnitudes, dogleg, crack shape and geomet-
Stress Intensity Factor rical parameters on the fatigue life. A detailed discussion of all these effects concludes the
S-N paper. It is the contention of the authors that the method can be easily, rapidly and successfully
Topic:
extended to other oilfield tubulars.
Stress-Number of Cycles to failure
© 2016 Elsevier Ltd. All rights reserved.
WF
Topic:
Weight Function

Nomenclature

a [m] Crack Size

a/c [-] Crack Aspect Ratio
a/t [-] Crack Length To Material Thickness Ratio
ai [m] Initial Crack Size
af [m] Final Crack Size
Δa [m] Crack Growth Increment
C Paris Material Constant
E [Pa] Material's Young Modulus
φ,Φ [°] Dogleg Angle
fm(x,a) 1/√m Weight Function

(continued on next page)

http://dx.doi.org/10.1016/j.engfailanal.2016.12.013
1350-6307/© 2016 Elsevier Ltd. All rights reserved.
12 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34

Y [-] Normalised Stress Intensity Factor

K [MPa√m] Stress Intensity Factor
ΔK [MPa√m] Stress Intensity Factor Range
M [N-m] Moment
m Paris Material Exponent
N [cycles] Number Of Cycles
Q [−] Elliptical Crack Shape Factor
R/t [−] Pipe Radius To Thickness Ratio
Δσi [MPa] Stress Range
σ [MPa] Stress
St [MPa] Newman-Raju's characteristic tensile plate's stress in the x-direction
Sb [MPa] Newman-Raju's characteristic anti-plane bending plate's stress in the x-direction

1. Introduction

Today's oil and gas wells are generally unconventional in nature, being either extended reach wells or ultra-deep wells. This
development is evident from the review of the drilling industry's evolution envelope [1,2], which is a graphical representation
of the records of geometries of wells drilled from across the globe. The review reveals a progressive shift and departure from ver-
tical well trajectories and geometries, which were previously the norm, to wells that are deeper and farther reaching.
Drilling deeper and farther reaching wells introduces a challenging fusion of complex trajectories and drillstring loading con-
ditions. Typical loading conditions involve the drillstring being under tension and bending stresses while undergoing rotation,
leading to increased fatigue of the drill-pipe system. Drillstring fatigue failures are the most common and costly types of failures
in the drilling industry [3] and contribute significantly to operational costs, loss of time and resources. Evaluating the fatigue life of
drilling pipes is imperative, considering that daily rates for hiring deep water rigs exceed $400,000.00 and the failures of
drillstrings in drilling service cost the oil industry $250 million annually, on average.
Determining the fatigue life of the drill pipes would require finding the number of cycles needed to propagate a fatigue crack
from an initial crack size ai [m] present in the drill-pipe material, to a final crack size, af [m] at which static failure occurs. The
fatigue life prediction through fatigue cracks growth depends on a reliable measurement or evaluation of the Stress Intensity Fac-
tor (hereafter referred to as SIF), which is the single most important crack tip characterisation parameter for any possible fatigue
life prediction by fatigue crack growth.
Several attempts have been made that utilise the Finite Element (FE) approach in evaluating Stress Intensity Factors (SIFs) for
several cylindrically shaped geometries, under different loading conditions, such as tension loadings, pure bending loadings, or
combined tension and bending loadings but not under the combined load combination of tension and bending while undergoing
rotation [4–6]. For drill pipes in drilling service only one attempt has been made [7]. However, the attempts based on the FE
method, inherently imply building new models and loading conditions every time any fatigue analysis is made, as well as require
a higher level of FE method skills to perform a fatigue evaluation. Therefore, owing to the immense computational resources avail-
able to the contemporary engineer, there is a need for strongly reliable numerical means of evaluating the SIF, different from the
FE method currently being used. This research presents an entirely numerical approach to drill pipe fatigue life evaluation,
utilising the Multiple Reference States Weight Function approach and an appropriate fatigue crack growth model. This is the orig-
inal contribution of the present paper.

2. Multiple reference states weight function technique

The Weight Function (WF) is a geometrical property of a cracked body, knowledge of which allows the determination of Stress
Intensity Factors (SIFs) for any given loading condition. The WF identifies a stress singularity in a cracked elastic domain as a func-
tion of the stress intensity at the crack tip independent of externally applied loads. The WF represents a means of separating the
load-geometry entanglement of a cracked body. Once a WF has been derived, it can be used to generate SIF's for different loading
conditions for the same geometry. Knowledge of stress distribution along a potential crack face and a WF for that geometry can be
extended for other loading conditions to generate SIFs. The WF is generally agreed to be the most efficient method for calculating
SIFs.
Rice [8] expressed the WF as:

H ∂U m ðx; aÞ
f m ðx; aÞ ¼ ð2:1Þ
2K m ∂a

where fm(x, a) is the WF, H an appropriate elastic constant, Km a known SIF, a the crack length, Um (x, a) the crack's opening dis-
placement, and x the distance from the mouth of the crack along the crack length.
The SIF can be expressed as:

a
K ¼ ∫ 0 σ ðxÞf m ðx; aÞdx 2:2
 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34 13

Fig. 1. Diagram showing the deepest and surface points (A and B) of a crack.

where σ(x) is the applied stress distribution on the crack face plane in an un-cracked specimen, and fm(x, a) is the WF.
Shen and Glinka [9] derived generalised forms of the WF at the surface and deepest points of a semi-elliptical surface crack in
a flat plate. For the deepest and surface points (A and B) of a crack as shown in Fig. 1, Shen and Glinka's WF can be expressed
as:

 
1  x  1  x   x  3
f m;A ðx; aÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 þ M 1A 1− a =2 þ M2A 1− a þ M3A 1− a =2 ð2:3Þ
2π ða−xÞ

 
1 x  1 x   x  3
f m;B ðx; aÞ ¼ pffiffiffiffiffi 1 þ M 1B a =2 þ M2B a þ M3B a =2 ð2:4Þ
πx

where x and a are defined above.

Eqs. (2.3) and (2.4) indicate that if the general form of the SIF is to be utilised for any given stress distribution, three coeffi-
cients M1, M2, and M3 would need to be determined for a cracked body. Determination of these coefficients is possible given a
knowledge of stress intensity factors for at least two reference load conditions, with the added condition from observations
that, under symmetrical loading conditions, the first and second derivatives of the crack surface of a central through edge crack
is 0 at x = 0, the crack tip [12].
This added condition is mathematically expressed as:

∂f m;A ðx; aÞ ∂f m;B ðx; aÞ

¼0 ¼0 ð2:5Þ
∂x ∂x

 
∂2 f m;A ðx; aÞ ∂2 f m;B ðx; aÞ
 ¼0  ¼0 ð2:6Þ
∂x2  ∂x2 
x¼0 x¼0

The expressions derived from Shen and Glinka's [9] work have been utilised by several researchers in the derivation of SIFs for
semi-elliptical surface cracks in cylinders of different thicknesses and dimensions. Wang and Lambert [10] utilised the WF tech-
nique in evaluating the SIFs for longitudinal semi-elliptical surface cracks, while Zheng et al. [11] applied the approach to evaluate
SIFs for internal surface cracks in thick walled cylinders. These results serve to demonstrate the reliability of the approach.
14 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34

For a given geometry and loading combination with at least two reference SIFs, the SIF could be written considering the fol-
lowing constitutive equations for the deepest point of the crack given in Eq. (2.3):
 
1  x  1  x   x  3
K r1 ¼ ∫ 0 σ r1 ðxÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 þ M 1A 1− a =2 þ M 2A 1− a þ M3A 1− a =2 dx
a
ð2:7Þ
2π ða−xÞ

 
1  x  1  x   x  3
K r2 ¼ ∫ 0 σ r2 ðxÞ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 þ M 1A 1− a =2 þ M 2A 1− a þ M3A 1− a =2 dx
a
ð2:8Þ
2π ða−xÞ

with the added condition that, from Eq. (2.5):

 !
∂ 1  x   x   x  
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 þ M1 1− a =2 þ M 2 1− a þ M3 1− a =2 
1 3
¼0 ð2:9Þ
∂x 2πða−xÞ 
x¼0

The three expressions represented by Eqs. (2.7) to (2.9) can be simultaneously solved to derive the three coefficients M1, M2,
and M3 that are unique to the given geometry under the specified loading conditions.
To improve accuracy and computational efficiency, Petroski and Ordjovic [13] proposed the Multiple Reference States (MRS)
technique that employs numerical integration instead of the numerical differentiation scheme of previous approaches, expressing
the numerical differentiation term as an arithmetic series and directly defining the crack opening displacement as an arithmetic
series:

∂U ða; xÞ 4σ 0 pffiffiffi m  x   j−1 =2

¼ 2 ∑ C j 1− a ð2:10Þ
∂a H j¼0

where Cj represents the coefficients of the WF, and all other symbols have been defined previously.

The initial constant is: C 0 ¼ 1 2 F 1 ða =t Þ and F 1 ða =t Þ ¼ σ Kp1 ffiffiffiffi for the first load case, where σ01 refers to the stress magnitude
0 1 πa
under this loading condition.
For any load case, Eq. (2.2) can be written as:

a H ∂U ðx; aÞ
∫0 σ ðxÞ m dx ¼ K i ðaÞK 1 ðaÞ ð2:11Þ
2 i ∂a

where K1(a) refers to the first loading condition, and i represents the i-th loading condition.
Substitution of Eq. (2.10) into Eq. (2.11) leads to:

pffiffiffi a
m  x   j−1 =2
2 2σ 0 ∫ 0 σ i ðxÞ ∑ C j 1− a dx ¼ K i ðaÞK 1 ðaÞ ð2:12Þ
j¼0

and if

a  x   j−1 =2
W ij ¼ ∫ 0 σ i ðxÞ 1− a dx ð2:13Þ

then
a   rffiffiffiffiffiffi
m F1 πa
∑ W ij C j ¼ K i ðaÞ t
ð2:14Þ
j¼0 2 2

This, with knowledge of C0, would give:

a   rffiffiffiffiffiffi !
m F1 πa
∑ W ij C j ¼ t
K i ðaÞ −W i0 : ð2:15Þ
j¼0 2 2

If the right hand side of Eq. (2.15) is taken as q, the summation of terms could be represented as a system of equations:

W 11 C 1 þ W 12 C 2 ¼ q1 ð2:16Þ

W 21 C 1 þ W 22 C 2 ¼ q2 ð2:17Þ
 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34 15

Using determinants:

 
q W 
 1 12 
q W 
C 1 ¼  2 22

 ð2:18Þ
 W 11 W 12 
W W 
21 22

 
q W 11 
− 1
q2 W 21 
C 2 ¼   ð2:19Þ
W
 11 W 12 
W W 22 
21

With knowledge of these coefficients, the WF is determined as

pffiffiffi
H ∂U ða; xÞ 2 2σ 0 m  x   j−1 =2
f m ða; xÞ ¼ ¼ ∑ C 1− a ð2:20Þ
2K 1 ðaÞ ∂a K 1 ðaÞ j¼0 j

This expression of the WF, from multiple reference states, represents a reliable and accurate means of evaluating the SIF.

3. Validation of MRS technique

The MRS approach was transformed into computer code and implemented in MATLAB. To guarantee confidence in the meth-
od, the code was progressively validated by using several geometries and loading conditions, and by using several published pa-
pers [10–12,14] as standards, some of which are mentioned below.

3.1. Edge crack in a finite width sheet

For a simple edge crack in a sheet of finite width, the reference solutions were obtained from collocation formulas of Brown
and Srawley in [15].
The reference solutions in normalised form are:

K a
a
2 a
3 a
4
pffiffiffiffiffiffi ¼ 1:12−0:23 þ 10:6 −21:7 þ 30:4 ð3:1Þ
σ 0 πa t t t t

Fig. 2. Pure bending loading [12].

16 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34

K a
a
2 a
3 a
4
pffiffiffiffiffiffi ¼ 1:12−1:39 þ 7:32 −13:1 þ 14:0 ð3:2Þ
6M πa t t t t
t2

For a =t ≤0:6.
The reference loading conditions in this case were pure tension and pure bending of the material; the loading distributions in
pure tension and pure bending are derived and expressed mathematically as:

σ tension ðxÞ ¼ σ 0 ð3:3Þ

σ bending ðxÞ ¼ σ 0 1− ð3:4Þ

NA

where NA represents the co-ordinate of the neutral axis.

The stress distributions are represented pictorially in Figs. 2 and 3 below.
The normalised reference tension and bending solutions were used as input to the WF script file, and the integrations were
performed numerically using a Lobatto Quadrature for numerical integration. The absolute error was set at 10−6 [-] and integra-
tions for the WF were performed by implementing the Lobatto WF and quadrature integration in a script m-file in MATLAB for
the same set of conditions as the reference case. The results obtained are shown in Fig. 4.
A statistical calculation of the coefficient of correlation between the collocation reference solutions and the MRS WF results
yielded a Pearson-Bravais correlation coefficient of 1 [-] for both the tension and bending cases respectively, indicating a highly
and positive correlation between the predictions of the two approaches, where the Pearson-Bravais correlation coefficient be-
tween any given pair of data sets x and y is defined as:

ðn∑xy−∑x∑yÞ
r xy ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi −1≤r ≤1 ð3:5Þ
n∑x2 −ð∑xÞ2 n∑y2 −ð∑yÞ2

Considering both the tension and bending loading conditions, the maximum % difference between the reference Normalised
SIF and the MRS WF derived Normalised SIF's was calculated as 0.9 [%]. Brennan [12] showed that the Petroski Achenbach tradi-
tional Crack Opening Displacement (COD) [16] method overestimated the reference solution from collocation formulae by 7 [%].
This demonstrates the high reliability of the MRS WF technique for accurately generating SIFs.

3.2. Semi-elliptical surface crack

The Newman and Raju [14] reference solutions were used to generate SIFs for the specific case of a semi-elliptical crack shape
shown in Table 1. The results obtained were compared with the result of Newman and Raju.

Fig. 3. Pure tension loading [12]. Po = σo is the magnitude of the far-field stress.
 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34 17

1.5
Reference
1.45 MRS WF

1.4

1.35

Normalized SIF
1.3

1.25

1.2

1.15

1.1

1.05

1
0 0.1 0.2 0.3 0.4 0.5
Normalized Crack Length (a/t)

Fig. 4. MRS WF Normalised SIF and reference solutions from collocation formulae [12].

For the sake of clarity, and to define the parametric angle, the cracks configuration used by Newman and Raju are shown in
Figs. 5 and 6 below.
Using the MRS WF technique, the following case was considered:where c/b [-] is the ratio of the half length of the crack to the
half width of the cracked plate.
The results generated using as inputs the parameters of Table 1 were plotted for validation (Figs. 7 and 8):
When plotted and compared with the Newman-Raju exact solutions, the results are shown to be accurate to within 0.1 [%].
With increasing confidence in the MRS results, it was possible to move on to more complex geometries and loading conditions
with emphasis on cylinders, the fundamental shapes for approximating drill-pipes.

3.3. Longitudinal semi-elliptical surface cracks in thin pipes.

Wang and Lambert [10] analytically computed weight functions and SIFs for internal and external surface cracks in pipes with
R/t ratio of 10 for the surface and at the deepest points of the crack employing as stress distributions:

x
n
σ ðxÞ ¼ σ 0 1− where n was set equal to 0; 1; 2; 3
a

The MRS WF approach was also used for the same set of load and boundary conditions; the results are presented for the
deepest and surface points for internal cracks of different aspect ratios, under parabolic and cubic loading conditions respectively
(Figs. 9 and 10):

3.4. Internal surface semi elliptical crack in thick walled cylinders

Zheng et al. [11] used an analytical WF approach to determine the SIFs for the surface and deepest points, A and B, of a cyl-
inder with Rint/t of 2, using load distributions of the power law form:

x
n 1 3
σ ðxÞ ¼ σ 0 where n was set equal to 0; ; 1; ; 2; 3 and σ 0 is the applied far−field stress:
a 2 2

The MRS WF approach was again used for the same set of load and boundary conditions, and the results are presented for the
deepest points for the crack and aspect ratios a/c of 0.2 and 1.0 (Figs. 11 and 12):
Having obtained excellent results from using the MRS Weight Function technique, it was believed that the confidence gained
could be extended to a drill-pipe in drilling service conditions.

Table 1
Input Parameters for Semi Elliptical Surface Crack Validation.

c/b a/c a/t Cases

0 0.20 1, 0.75, 0.5,0 Tension/bending

18 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34

Fig. 5. Semi elliptical surface crack [14].

4. Drill-pipe in drilling service conditions

A drill-pipe in drilling service passing through a dogleg is under the action of several stresses induced from rotation, bending
and the tensile loading on the drill-string. The coupling of several forces leads to fatigue of the drill-pipe material. For this study,
the drill-pipe was approximated as a hollow cylinder with geometry RRext int
¼ 1:1, (the ratio of the external to internal radii of the
hollow cylinder) passing through a curved section of a well while under rotation. This scenario results in complex loading of
the drill-pipe material, represented by combined tension and bending under rotation through an angle φ. The approach developed
here can be extended to various geometries and crack shapes, but in this study it is limited to an internal circumferential semi-
elliptical surface crack, as shown in Fig. 13.
For the drill-pipe shown in Fig. 13, the direction of crack growth is such that it is perpendicular to the cylinder axis which is
also the principal stress direction. As in most crack propagation analysis, the crack was assumed to be semi-elliptical,
characterised by the dimensionless parameters a/c [-], the crack aspect ratio and a/t [-] the relative crack depth. Fig. 14 shows
cracks of various aspect ratios.

4.1. Description and mathematical formulation

The cylinder of Fig. 13 is subjected to a tensile load T [N] and its rotation produces a bending moment, M [N-m] which can be
split into its constituent components in the x- and y-directions (x-y plane):

Mx ¼ M cosφ and M y ¼ M sinφ ð4:1Þ

Newman and Raju [14] expressed the stress intensity factor for a surface crack in a flat plate under combined load conditions
of tension and bending as:
rffiffiffiffiffiffiffiffi
a a a c

K I ¼ ðSt þ HSb Þ π F ; ; ; φ ð4:2Þ

Q t c b

Fig. 6. Coordinate system used to define parametric angle [14].

 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34 19

1.8

1.6

1.4

1.2

a/Q
1
Kr/Sr WF a/t = 0
0.8 WF a/t = 0.5
WF a/t = 0.75
0.6
WF a/t = 1
Newman a/t = 0
0.4
Newman a/t = 0.5
Newman a/t = 0.75
0.2
Newman a/t = 1
0
0 0.2 0.4 0.6 0.8 1
2 /

Fig. 7. MRS Normalised SIF's and Newman-Raju's reference solutions for a/c = 0.2, semi-elliptical surface crack in tension loading conditions, [14].

where St [MPa] is the characteristic tensile plate's stress and Sb [MPa] is the characteristic anti-plane bending plate's stress in the
x-direction, and the other parameters H, Q and F derive their basis from engineering judgement. This was extended to the drill-
pipe being considered.
A strong property of the SIF is that it lends itself to superposition; the superposition property was used to combine the mode I
SIF for the tension and bending cases in order to generate an overall expression for the drill-pipes SIF,

K I;overall ¼ K I;Tension þ K I;Bending x þ K I;Bending y ð4:3Þ

The bending moment vector has been resolved into components in the X and Y directions.
The SIF can be written in general form for the cylinder geometry as:

rffiffiffiffiffiffiffiffi
a
K ¼ Yσ π ð4:4Þ
Q

where Y [-] is the dimensionless Y-factor solution, and all other symbols have been defined previously.

1.5

1
a/Q
Kr/Sr

0.5
WF a/t = 0
WF a/t = 0.5
WF a/t = 1
Newman a/t = 0
Newman a/t = 0.5
Newman a/t = 1
0
0 0.2 0.4 0.6 0.8 1
2 /

Fig. 8. MRS Normalised SIF's and Newman-Raju's reference solutions for a/c = 0.2, semi-elliptical surface crack in bending loading conditions, [14].
20 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34

o Wang and Lambert's a/c=0

MRS WF a/c=0.05
2.5 a/c=0.1
a/c=0.2
a/c=0.4
2 a/c=1

Normalized SIF[-]
1.5

0.5

0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
a/t[-]

Fig. 9. MRS WF Normalised SIFs with Wang and Lambert's result for the deepest point of an internal longitudinal, semi elliptical surface crack, [10], (different as-
pect ratios).

The individual components of the SIFs can be written in form of Eq. (4.4) as:

rffiffiffiffiffiffiffiffi
a
K I;Tension ¼ Y I;Tension σ Tension π ð4:5aÞ
Q

rffiffiffiffiffiffiffiffi
a
K I;bending x ¼ Y I;Bending x σ bending x π ð4:5bÞ
Q

rffiffiffiffiffiffiffiffi
a
K I;bendingy ¼ Y I;Bending y σ bending y π ð4:5cÞ
Q

1
a/c=0.05
o Wang and Lambert's a/c=0.1
0.9
MRS WF a/c=0.2
a/c=0.4
0.8 a/c=1
Normalized SIF[-]

0.7

0.6

0.5

0.4

0.3

0.2
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
a/t[-]

Fig. 10. MRS WF Normalised SIFs with Wang and Lambert's result for the surface point of an internal longitudinal, semi elliptical surface crack, [10] (different as-
pect ratios).
 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34 21

3
pure-tension O Zheng et al
pure-bending MRS WF
2.5 n=0
n=1
n=2
n=3
2
n=1/2

a/Q [ - ]
n=3/2

1.5
Kr/Sr

0.5

0
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
a/t [ - ]

Fig. 11. MRS WF Normalised SIFs with Zheng et al.'s result for the deepest point,-internal crack, R/t = 2, a/c = 0, [11].

which can be combined into an Overall SIF:

h irffiffiffiffiffiffiffi
a
ffi
K I;overall ¼ Y I;Tension σ Tension þ Y I;Bending x σ bending x þ Y I;Bending y σ bending y π ð4:6Þ
Q

The bending stresses can be obtained from the generalised bending equation for the components of the bending stresses in the
X and Y directions:

M x Rext M y Rext M Rext

σ bending x ¼ ; σ bending y ¼ and σ bending ¼ ð4:7Þ
I I I

where Mx and My are defined in Eq. (4.1).

The overall SIF of the drill-pipe was derived as

h irffiffiffiffiffiffiffi
a
ffi
K I;overall ¼ Y I;Tension σ Tension þ Y I;Bending x σ bending cosφ þ Y I;Bending y σ bending sinφ π ð4:8Þ
Q

1.3
pure-tension O Zheng et al
pure-bending MRS WF
1.2
n=0
n=1
1.1
n=2
n=3
1 n=1/2
a/Q [ - ]

n=3/2
0.9
Kr/Sr

0.8

0.7

0.6

0.5
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
a/t [ - ]

Fig. 12. MRS WF Normalised SIFs with Zheng et al.'s result for the deepest point,-internal crack, R/t = 2, a/c = 1.0, [11].
22 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34

Fig. 13. Cracked cylinder loading and decomposition of the bending moment [7].

The reference stress intensity factors for pure tension and pure bending stress distributions for an internal circumferential
semi-elliptical surface crack in a cylinder for the surface and deepest points were taken from [9] for RRext
int
¼ 1:1, and condensed
below
Deepest point, A:

 a
a
2 rffiffiffiffiffiffiffi
a
ffi
K I;tension;A ¼ σ tension A0 þ A1 þ A2 π ð4:9Þ
t t Q

 a
a
2 rffiffiffiffiffiffiffi
a
ffi
K I;bending;A ¼ σ bending B0 þ B1 þ B2 π ð4:10Þ
t t Q

Surface Point, B:

h a
c1 irffiffiffiffiffiffiffi
a
ffi
K I;tension;B ¼ σ tension C 0 π ð4:11Þ
c Q


rffiffiffiffiffiffiffiffi
a D1 a
K I;bending;B ¼ σ bending D0 π ð4:12Þ
c Q

where A,B-,C and D– are all functions of the crack's geometrical parameters a/t [-], a/c [-].
The equations above are for the ranges 0:1≤ at ≤0:8 and 0:2≤ ac ≤1:0 where all symbols have their already defined meanings.

Fig. 14. Crack aspect ratios and shapes.

 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34 23

The multiple reference states approach was applied individually to the surface and deepest points of the crack, using the given
SIFs of Eqs. (4.9)–(4.12) in tension and bending as references leading to the derivation of the WF of Eq. (2.20).

pffiffiffi
H ∂U ða; xÞ 2 2σ 0 m  x   j−1 =2
f m ða; xÞ ¼ ¼ ∑ C 1− a ð2:20Þ
2K 1 ðaÞ ∂a K 1 ðaÞ j¼0 j

A combination of the WF and the derived stress distributions enables the determination of the SIF expressed in normalised
form as Y factor solutions (presented in the Results and discussions section).

4.2. Fatigue crack growth

The Paris fatigue crack growth model was applied in this study, the Paris equation represents the relationship between the
Fatigue crack growth rate and the stress intensity factor range, given as:

da m
¼ CΔK ð4:13Þ
dN

where

pffiffiffiffiffiffi
ΔK ¼ Y ðΔσ Þ πa ¼ KI; max−KI; min ð4:14Þ

Parameters C and m are experimentally derived and, for the purpose of application, should be determined for the particular
drill-pipe material considered. The Paris equation was applied locally to the crack as the crack advances, and the crack was as-
sumed to remain semi-elliptical at all times during the crack growth; two points are necessary to describe the propagation behav-
iour, taken to be A, the deepest point of the crack, and C, a generic point on the advancing front, as seen in Fig. 15:
Ngoc and Sellami [7] showed that the choice of point C does not have considerable effect on the fatigue life of the drill-pipe;
hence Point C was selected arbitrarily and the crack was assumed to satisfy the ellipse equation at all times i.e.:

x2 y2
þ ¼1 ð4:15Þ
a2 b2

This crack growth model was used to predict the remaining life of the drill-pipe specimen. A curve could be fitted to generate
a function Y = f(a) but in this case, due to the constant amplitude stress range, the SIF is used for simplicity. The technique used
here is to grow the crack in discrete amounts, i.e. discretise the crack lengths into relatively small amounts, so as to minimise the
crack increment da[m]. Using this approach, a constant K [MPa√m] is assumed for each crack increment da[m] and is updated to a
new K [MPa√m] for every new increment.

X
R

A
t

C B
b

Fig. 15. Crack shape modeling.

24 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34

From Eq. (4.13):

da
dN ¼ ð4:16Þ
CΔK m

and:

aiþ1 −ai
Niþ1 −Ni ¼ ð4:17Þ
CΔK m

The discretisation is chosen to be sufficiently small and an expression for the maximum and minimum value of the SIF is given
by Ngoc and Hedi [7].

 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffirffiffiffiffiffiffiffiffi
a
K I; max ¼ Y I;tension σ tension þ σ bending Y I;bending x 2 þ Y I;bending y 2 π ð4:18Þ
Q

 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffirffiffiffiffiffiffiffiffi
a
K I; min ¼ Y I;tension σ tension −σ bending Y I;bending x 2 þ Y I;bending y 2 π ð4:19Þ
Q

When KI,min [MPa√m] is negative, the compressive stresses do not contribute to crack propagation so that KI,min [MPa√m] is
usually set equal to zero and the SIF range (Eq. (4.14) is therefore reduced. Moreover, it should be noted that the presence of
a plastic zone ahead the crack tip implies the raising of the crack closure effect, related to negative stresses in the plastic wake,
which will further reduce the SIF range to be used in Eq. (4.13). For the sake of simplicity, the present analysis will not consider
the crack closure effect, which needs experimental data to be properly taken into account. Observe that this choice leads to (al-
though less precise) conservative predictions.
Eq. (4.17) implies that the present life of the material can be derived from the conditions in the previous state, and the errors
associated with this approach are known to decrease significantly with the magnitude of the discretisation of the crack length in-
crement, i.e. the smaller the increment in crack length, the higher the accuracy of the results under this assumption.

5. Results and discussions

The models developed in Section 3 and the MRS technique were transformed into computer codes for simulation and para-
metric studies; first the SIF was evaluated and the results used as input for generating the fatigue crack growth curves. The influ-
ence of several parameters on the drill-pipe fatigue life was studied independently including:
• Effect of the crack geometry
• Effect of angle of bending
• Effect of increasing tension
• Effect of increasing bending moments
• Effect of assumed initial flaw size
The S-N (Stress-Number of cycles to failure) curves were then obtained using the two stress components (tensile and bending)
while keeping the tensile stress constant and studying the influence of the cyclic bending stress.

5.1. Calculation of the Normalised SIF's

The base case input data are given in Table 2:

Table 2
Arbitrary input data for the drill-pipe.

Quantity Symbol Value Unit

1 Applied tensile stress σtension 50 [MPa]

2 Applied bending stress σbending 150 [MPa]
3 Bending angle Φ 15 [deg]
4 Crack aspect ratio a/c 0.2 [-]
5 Crack relative depth a/t 0.1 [-]
6 Surface point – 0 [rad]
7 Deepest point – π/2 [rad]
 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34 25

plot of non dimensional SIF vs.crack depth to plate thickness ratio

1.5

pure-tension
pure-bending
WF Tension
1
WF bending

a/Q [ - ]
Bending 1
Bending 2
Combined Bending
r
K /S
r

0.5

0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
a/t [ - ]

Fig. 16. Deepest point SIF, for bending angle of 0.

The Normalised SIF was expressed with the following relation in all cases:

Kr
Y¼ rffiffiffiffiffiffiffi
ffi ð5:1Þ
a
Sr π
Q

where Y [-] is the normalised SIF and Kr [MPa√m], Sr [MPa] are reference SIF and tensile stress respectively.
The implementation of the MRS WF in MATLAB, using the reference tension and bending solutions (for the deepest and sur-
face point) of the crack [9], generates for a straight section (zero dogleg) Figs. 16 and 17 respectively.
The results predict that the SIF for bending degenerates into the pure bending solution for a vertical drill-pipe rotating in a
straight section. The simulation was repeated for the base case where the dogleg is 15° and Figs. 18 and 19 are obtained for
the surface and deepest points of the crack respectively.
From the curves it can be seen that a long crack with aspect ratio 0.2 [-] will have far higher stress intensities at the deepest
point than at the surface point under the tension loadings, while it will have relatively similar SIF for bending loadings. A qual-
itative inspection of the surface point also reveals some points of inflexion, indicative of a constant SIF over the range 0.3 b a/
t b 0.5. The results also indicate that the combined effect of the SIF in bending (Φ = 15°) is greater than it would have been
in a pure bending scenario (Φ = 0°), implying that increasing the bending angle Φ amplifies the intensity of the stresses in an

plot of non dimensional SIF vs.crack depth to plate thickness ratio

0.9

0.8

0.7

0.6
a/Q [ - ]

0.5
pure-tension
pure-bending
0.4
Kr/Sr

WF Tension
WF bending
0.3 Bending 1
Bending 2
0.2 Combined Bending

0.1

0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
a/t [ - ]

Fig. 17. Surface point SIF, for bending angle of 0.

26 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34

plot of non dimensional SIF vs.crack depth to plate thickness ratio

0.9

0.8

0.7

0.6

a/Q [ - ]
0.5
Kr/Sr pure-tension
pure-bending
0.4 WF Tension
WF bending
Bending 1
0.3
Bending 2
Combined Bending
0.2

0.1
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
a/t [ - ]

Fig. 18. Surface point SIF, for bending angle of 15°.

existing crack tip. For the surface point, this even exceeds the intensity of the stresses in tension conditions, implying that the
bending effect contributes more to fatigue damaging.
It was necessary to generate results for a/t N 0.8: The results below were obtained for the surface and deepest points
respectively.
From Fig. 20, for the surface point, beyond a/t of 0.8 the solution becomes unstable, since large stress gradients are encoun-
tered. A possible reason is that the linear elastic region, where the Paris law is valid, has been exceeded. The deepest point still
maintains expected trends as seen in Fig. 21, but it can be ruled out that the analysis could be best done for a/t ≤ 0.8, making
the surface point the limiting point for the analysis (in terms of the crack length and validity of the reference equations).
Hence for the rest of this study failure was defined as the point where the crack reaches 0.8 [-] of the thickness of the drill-
pipe since beyond this the present model being used would lose its validity and any predictions made would likely be unreliable.

5.2. Influence of crack shape on SIF

The effect of changing the shape of the crack on the intensities of the stresses at the crack tip was studied by varying the crack
aspect ratio from 0 to 1; the result for the surface and deepest points are shown below (Fig. 22):

plot of non dimensional SIF vs.crack depth to plate thickness ratio

1.5
pure-tension
pure-bending
WF Tension
WF bending
Bending 1
Bending 2
1 Combined Bending
a/Q [ - ]
Kr/S r

0.5

0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
a/t [ - ]

Fig. 19. Deepest point SIF, for bending angle of 15°.

 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34 27

plot of non dimensional SIF vs.crack depth to plate thickness ratio

0.9

0.8

0.7

0.6

a/Q [ - ]
0.5 pure-tension
pure-bending
Kr/Sr
0.4 WF Tension
WF bending
0.3 Bending 1
Bending 2
0.2 Combined Bending

0.1

0
0 0.2 0.4 0.6 0.8 1
a/t [ - ]

Fig. 20. Surface point SIF, for bending angle of 15°, to 0 ≤ a/t ≤ 1.

In both the cases of surface and deepest points, the SIF gradients are positive, indicating that as the crack gets longer, the SIF
increases, and hence the crack propagation will be expected to get faster; it is also evident that in all cases the combined bending
SIF is greater than the SIF in pure bending.

5.3. Fatigue life

The fatigue crack growths described by Eq. (4.17) was implemented in a MATLAB Script file and for the base condition given in
Table 2 the following results were obtained.
The crack growth rate da/dN can be represented as a tangent to the curves at any point. And for a particular curve, it shows
that the gradient starts from relatively small values and then begins to grow faster. This implies that as a crack grows larger, the
intensities of the stresses at the cracks tip increase and hence the speed of the propagation increases. From the results in Figs. 23
and 24, the deepest point proves to be more critical for the base case aspect ratio of a/c = 0.2 since here it predicts a shorter life
for the drill-pipe, less than 65,000 cycles compared to the surface point prediction of about 90,000 cycles. Correspondingly from
here onwards all evaluations are done for the deepest point since this would have caused fatigue failure at a fewer number of
cycles compared with the surface point of the crack.

plot of non dimensional SIF vs.crack depth to plate thickness ratio

1.8
pure-tension
pure-bending
1.6
WF Tension
WF bending
1.4 Bending 1
Bending 2
1.2 Combined Bending
a/Q [ - ]

0.8
r
K /S
r

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1
a/t [ - ]

Fig. 21. Deepest point SIF, for bending angle of 15°, to 0 ≤ a/t ≤ 1.
28 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34

Surface Point Deepest Point

1.4 2.6
0
2.4 0.1
1.2
0.2
2.2 0.3
1 0 0.4
2 0.5

Normalized SIF [ - ]
Normalized SIF [ - ]

0.1
0.2 0.6
0.8 1.8
0.3 0.7
0.4 0.8
0.6 1.6 0.9
0.5
0.6 1
1.4
0.4 0.7
0.8 1.2
0.9
0.2
1 1

0 0.8
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
a/t [ - ] a/t [ - ]

(a) (b)
Surface Point Deepest Point
1.4 1.2 0
0 0.1
0.1 1.1
1.2 0.2
0.2 1 0.3
0.3 0.4
1 0.4 0.9 0.5
Normalized SIF [ - ]
Normalized SIF [ - ]

0.5 0.6
0.6 0.8
0.8 0.7
0.7 0.8
0.7
0.8 0.9
0.6 0.9 0.6 1
1
0.4 0.5

0.4
0.2
0.3

0 0.2
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
a/t [ - ] a/t [ - ]

(c) (d)
Surface Point Deepest Point
1.4 1.3
0
1.2 0.1
1.2
0.2
1.1
0.3
1 1 0.4
0
Normalized SIF [ - ]

Normalized SIF [ - ]

0.5
0.1 0.9
0.8 0.6
0.2
0.7
0.3 0.8
0.8
0.6 0.4
0.7 0.9
0.5
1
0.6
0.4 0.6
0.7
0.8 0.5
0.2 0.9
1 0.4

0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
a/t [ - ] a/t [ - ]

(e) (f)
Fig. 22. Variation of SIF with aspect ratio (a/c) for the surface and deepest points, in pure tension (a, b), bending (c, d) and combined bending (e, f).
 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34 29

-3
x 10
8

crack length, a[m]

5

0
0 2 4 6 8 10
Number of Cycles 4
x 10

Fig. 23. Crack growth curve for the base case, surface point.

The developed MATLAB code was run for a very wide range of conditions to ascertain the influence of the following param-
eters on the number of cycles to failure for the drill-pipe.

5.3.1. Effect of crack shape and geometry

The data for the base case in Table 2 was used and the crack aspect ratio was varied over the range 0 ≤ a/c ≤ 1: The results are
shown in Fig. 25.

-3
x 10
8

6
crack length, a[m]

0
0 1 2 3 4 5 6 7
Number of Cycles x 10
4

Fig. 24. Crack growth curve for the base case, deepest point.
30 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34

-3
x 10
8

crack length, a[m]

a/c = 0
5
a/c =0.1
4 a/c =0.2
a/c =0.3
3 a/c =0.4
a/c =0.5
2 a/c =0.6
a/c =0.7
1 a/c =0.8
a/c =0.9
0 a/c =1 6
0 1 2 3 4 5
Number of cycles 5
x 10

Fig. 25. Influence of crack shape on crack growth curve.

From the results it can be said that the crack shape influences the fatigue life of the drill-pipe and the observed trend is that a
longer crack, e.g. a/c = 0.1, would fail before a shorter crack of e.g. a/c = 0.8. In general terms, a larger crack is expected to fail
before a smaller crack.

A qualitative assessment can also reveal that the gradient or crack growth rate da dNwill be gentle, or slower for shorter cracks,
and faster for longer cracks.

5.3.2. Effect of the dogleg

The angle of bending was varied from 0° ≤ Φ ≤ 30° and results obtained are shown in Fig. 26 below.
From Fig. 26, it can be seen that bending the drill-pipe through an angle Φ, contributes to a significant reduction in its fatigue

life. The crack growth rate da dN increases with the angle Φ, as evidenced in the slopes of the crack growth curves; hence it is
advisable to avoid putting a rotating drill-pipe through large deviations or severe doglegs, as this contributes significantly to fa-
tigue life reduction.

5.3.3. Effect of increasing tension

The effect of the tension on the drill-pipe was investigated considering the following interval, 20 [MPa] ≤ σtension ≤ 120 [MPa]:
The results are shown in Fig. 27 below.
Increasing tensile stresses is seen to shorten the drill-pipe's fatigue life, because the tensile stresses are known to be detrimen-
tal and contribute significantly to KI [MPa√m] increases. When the tensile stresses increase, there is a magnification of the stress
singularity at the crack tip which leads to faster crack propagation and shorter fatigue life of the drill-pipe. Hence the tensile loads
should be made to be as low as reasonably possible, possibly by buoying the hook load using appropriate drilling mud.

5.3.4. Effect of bending stress

By varying the cyclic bending stresses over the following interval of stresses 90[MPa]≤ σBending ≤ 190[MPa], the results shown in
Fig. 28 were obtained, demonstrating that with increasing bending stresses the fatigue life of the material is reduced.
da
In terms of the crack growth rate, a qualitative inspection of the curves would reveal that the crack growth rate dN vsΔK curve
represented by using tangents on the curve of the fatigue life would yield higher magnitudes with increasing rates. This implies
-3
x 10
8
=0
7 =3
=6
6
=9
crack length, a[m]

5 = 12
= 15
4 = 18
= 21
3
= 24
2 = 27
= 30
1

0
0 1 2 3 4 5 6 7 8 9 10
Number of cycles 4
x 10

Fig. 26. Influence of dogleg on crack growth curve.

 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34 31

-3
x 10
8

6
20[MPa]

crack length, a[m]

5 30[MPa]
40[MPa]
4 50[MPa]
60[MPa]
3 70[MPa]
80[MPa]
2 90[MPa]
100[MPa]
1 110[MPa]
120[MPa]
0
0 2 4 6 8 10 12 14
Number of cycles 4
x 10

Fig. 27. Effect of tensile stress on the crack growth.

that increasing the bending load increases the Fatigue Crack Growth Rate (FCGR) and decreases the fatigue life of the material;
conversely, decreasing the load would reduce the FCGR and increase the fatigue life of the material. Consequently, drill-string vi-
bration should be minimised during the drilling operations so as to reduce the cyclic stresses.

5.3.5. Effect of initial flaw size

In the previous analyses we have assumed an initial relative crack depth equal to 0.1. The initial flaw size is very important in
the field of fracture mechanics because every analysis begins with the assumption that flaws of a particular size exist in the ma-
terial, and every FCG prediction is made relative to the initial flaw size assumed. This parameter was varied with initial crack sizes
of [0.9 mm, 1.8 mm, 2.7 mm and 3.6 mm] respectively, corresponding to a relative crack depth a/t [-] equal to 0.1, 0.2, 0.3, 0.4, to
investigate the initial crack size's influence on the fatigue life of the material: The results are shown in Fig. 29.
A qualitative inspection of the curves would predict steeper gradients for larger cracks and gentle gradients for smaller cracks.
The engineering implication is that for a constant loading, an initially large flaw would be expected to propagate faster giving a
shorter fatigue life; conversely, a smaller initial flaw would propagate slower yielding a longer life, and for travelling on the same
fatigue life curve the crack growth rate would be seen to increase with the crack size at that point: As the crack grows in size, the
rate of its growth would be expected to increase in proportion to its size (depth in this case) at that instant. From the gentle to
steep transition of the slopes of the generated curves, it can be seen that a large part of the crack growth life is spent initiating
and propagating the short cracks; hence identification and sizing of the initial flaw is important for reliable estimation of the fa-
tigue life of the drill-pipe material.
It is necessary to stress that in our analysis the initial crack size is considered as an input parameter (strongly affecting the
component fatigue life) since it's assumed that cracks are detected during inspections. To determine the fatigue life of an intact
drill pipe, it is necessary to take care of the initiation phase and of the short fatigue crack growth. However, this issue goes beyond
the scope of the present paper and will not be considered here.

5.4. Stress-number of cycles to failure (S-N) curve

A plot of the cyclic bending stresses was made against the number of cycles to failure, as shown in Fig. 30. The SN curve is a
graphic depiction of the number of cycles the drill-pipe can take until failure for a particular value of the applied stress.

-3
x 10
8

6
crack length, a[m]

90[MPa]
5
100[MPa]
110[MPa]
4
120[MPa]
3 130[MPa]
140[MPa]
2 150[MPa]
160[MPa]
1 170[MPa]
180[MPa]
0 190[MPa]
0 5 10 15
Number of cycles 4
x 10

Fig. 28. Effect of bending stress on crack growth.

32 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34

-3
x 10
8
a/t= 0.1
7 a/t=0.2
a/t=0.3
6 a/t=0.4

crack length, a[m]

5

0
0 1 2 3 4 5 6 7
Number of cycles 4
x 10

Fig. 29. Effect of initial flaw size on crack growth.

On the basis of what observed in the previous sections, the number of cycles to failure is determined assuming:

(i) the Paris' law holds;

(ii) the elliptical crack shape is assumed to remain constant during crack growth;
(iii) the drill pipe fails when the relative crack depth a/t [-] reaches the threshold value of 0.8.

As the curve progresses, it becomes asymptotic to the number of cycles to failure axis because it would take an infinite number
of cycles for the material to fail when the applied stress on the material is zero. This SN curve can prove to be a very helpful tool
from a stress–life analysis standpoint.

5.4.1. Influence of varying crack shape, initial flaw size and bending angle
SN curves were generated for various conditions, to identify trends on changing the previously-mentioned parameters. Fur-
thermore it has been observed that generally a relative shift to the right on the S-N space indicates an improvement in fatigue
life, and a relative shift to the left indicates a reduction in fatigue life. Hence from Figs. 31 to 32, it can be concluded that an in-
crease in crack aspect ratio leads to significant increases in the fatigue life of the drill-pipe and vice-versa, increasing dogleg re-
duces fatigue life and vice-versa, and a larger starting crack reduces fatigue life significantly and vice-versa (Fig. 33).
The gradients also demonstrate the gains in fatigue life resulting from small changes in stresses under fixed conditions of, for
example, initial flaw size, crack shape or dogleg. Considering the curves for the crack shape, it can be inferred that for the shorter
crack, a =c ¼ 0:6, there will be relatively greater improvement in fatigue life by reducing the cyclic stresses compared to a longer
crack, a =c ¼ 0:2.

200
Base Case

180
Bending Stress [MPa]

160

140

120

100

80
2 4 6 8 10 12 14 16
Number of Cycles x 10
4

Fig. 30. S-N curve for the base case.

 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34 33

200
a/c = Base
a/c = 0.4
a/c = 0.6
180

b [MPa]
160

Bending stress,
140

120

100

80
0 1 2 3 4 5
Number of cycles x 10
5

Fig. 31. Influence of crack shape on SN curve.

6. Conclusions

In this paper the multiple reference state weight function method has been used to determine the stress intensity factor for an
internal circumferential semi elliptical surface crack in a drill-pipe in drilling service conditions, for combined loads of tension and
bending under rotation. This approach represents a rapid, easily implementable and accurate numerical means of quantitatively
evaluating SIF for single and multi-dimensional cracks.
Fatigue crack growth of a drill-pipe is evaluated using the Paris Law. The simulations performed reveal the nature of influence
on the SIF and drill-pipe fatigue life, of the loading and geometrical parameters.
The simulations show that putting a drill-pipe through a dogleg, increases the overall SIF. Furthermore, increased bending
under rotation significantly reduces drill-pipes fatigue life by accelerating fatigue crack growth. The influence of the tensile
loads which are detrimental to the drill-pipes fatigue life by accelerating crack growth is also seen. Moreover, a larger assumed
starting crack accelerates fatigue failure. Finally, the crack shape and geometrical parameters are also seen to significantly influ-
ence fatigue life.
A careful combination or control of the influencing parameters can aid in drill-pipe fatigue life extension and improve decision
making, resulting in significant savings in cost and time. In fact, useful drill-pipes can remain in use for a longer time than cur-
rently the practice. The approach presented in this paper can be easily extended to specific applications where detailed informa-
tion such as cylinder dimensions, crack type, crack position and load distributions is available. The findings in this study should be
of great utility in the drilling industry.

200
a/t = Base Case
a/t = 0.2
180 a/t = 0.3
b [MPa]

160
Bending stress,

140

120

100

80
0 0.5 1 1.5 2
Number of cycles x 10
5

Fig. 32. Influence of initial flaw size on SN curve.

34 C. Ojanomare et al. / Engineering Failure Analysis 74 (2017) 11–34

200
=0
Base Case
180 = 30

b [MPa]
160

Bending stress,
140

120

100

80
0 0.5 1 1.5 2
Number of cycles 5
x 10

Fig. 33. Influence of dogleg on SN curve.

References

[1] R. Thomas, J. McCormick, Longer deviated wells push drill pipe limits. accessed at http://www.drillingcontractor.org/longer-deviated-wells-push-drill-pipe-
limits-8779 2011 (on June 3rd 2015).
[2] http://www.drillingcontractor.org/continuous-improvements-lead-to-maersk-oil-qatars-longest-horizontal-well-in-the-world-1924 (accessed on December
3rd 2016).
[3] http://petrowiki.org/Drillpipe_failures (assessed on 2nd March 2015).
[4] A. Carpinteri, R. Brighenti, A. Spagnoli, Fatigue growth simulation of part-through flaws in thick-walled pipes under rotary bending, Int. J. Fatigue 22 (1) (2000)
1–9.
[5] A. Carpinteri, R. Brighenti, A. Spagnoli, Part-through cracks in pipes under cyclic bending, Nucl. Eng. Des. 185 (1) (1998) 1–10.
[6] A. Carpinteri, R. Brighenti, Part-through cracks in round bars under cyclic combined axial and bending loading, Int. J. Fatigue 18 (1) (1996) 33–39.
[7] H.D. Ngoc, S. Hedi, Stress intensity factors and fatigue growth of a surface crack in a drill pipe during rotary drilling operation, Eng. Fract. Mech. 96 (2012)
626–640.
[8] J.R. Rice, Some remarks on elastic crack tip stress fields, Int. J. Solids Struct. 8 (1972) 751–758.
[9] G. Glinka, Development of weight functions and computer integration procedures for calculating stress intensity factors around cracks subjected to complex
stress fields, Prepared on Behalf of SaFFD Inc.,Ontario, 1996.
[10] X. Wang, S.B. Lambert, Stress intensity factors and weight functions for longitudinal semi-elliptical surface cracks in thin pipes, Int. J. Press. Vessel. Pip. 65 (1996)
75–87.
[11] X.J. Zheng, A.C. Kiciak, G. Glinka, Weight functions and stress intensity factors for internal surface semi-elliptical crack in thick walled cylinder, Eng. Fract. Mech.
58 (3) (1997) 207–221.
[12] F.P. Brennan, Evaluation of stress intensity factors by multiple reference state weight function approach, J. Theoret. Appl. Fract. Mech. 20 (1994) 249–256.
[13] R.P. Ojdrovic, H.J. Petroski, Weight functions from multiple reference states and crack profile derivatives, Eng. Fract. Mech. 39 (1991) 105–111.
[14] J.C. Newman, I.S. Raju, An empirical stress intensity factor equation for the surface crack, Eng. Fract. Mech. 15 (1) (1981) 185–192.
[15] D.J. Cartwright, D.P. Rooke, Approximate stress intensity factors compounded from known solutions, Eng. Fract. Mech. 6 (1974) 563–571.
[16] H.J. Petroski, J.D. Achenbach, Computation of the weight function from a stress intensity factor, Eng. Fract. Mech. 10 (1978) 257–266.