of 35
7/28/2019 2006_2p
1/35
1/35
Fatigue on drill string conicalthreaded connections,
test results and simulations
A. Baryshnikov
L. Bertini,
M. Beghini,
C. SantusENI S.p.A.
Milano.
Italy
University of Pisa,
mechanical dept.
Italy
7/28/2019 2006_2p
2/35
2/35
Short introduction to drilling technology- drill string and drill pipes, fatigue failures on drill pipes
- steel heavy construction vs. aluminum light construction
Full scale fatigue tests- description of test rigs
- test results
Finite Element simulations- FE model dedicated to threaded connection
Fatigue models- classic approach (Gerber, kf, surface effect)- test results correlation
Conclusions
Contents
7/28/2019 2006_2p
3/35
3/35
Short introduction to drilling technologydrill string and drill pipes, fatigue failures on drill pipes
Drill String
7/28/2019 2006_2p
4/35
4/35
Short introduction to drilling technology
Drill string:
hundreds ofdrill pipes
connected through threaded
connections
Drill pipe length ~ 10m
Drill string max. length ~ 5km
Basic nomenclature
Dog leg segment, for
deviated drilling
Drill bit
drill string and drill pipes, fatigue failures on drill pipes
7/28/2019 2006_2p
5/35
5/35
Short introduction to drilling technologyFatigue locations along drill string
Rotating bending fatigue,
due to dogleg on the upper
part of the string
Lateral bending fatigue, due
to buckling on the lower part
of the string
drill string and drill pipes, fatigue failures on drill pipes
7/28/2019 2006_2p
6/35
6/35
Fatigue locations along drill string
Fatigue accounts for 70 % of failures
Corrosion, Stress-Corrosion, Wear, Static stresses are further
detrimental effects in combination with fatigue
Short introduction to drilling technologydrill string and drill pipes, fatigue failures on drill pipes
7/28/2019 2006_2p
7/35
7/35
Short introduction to drilling technology
Steel construction Aluminum construction
Aluminum
body pipe
Steel thread
connection
(tool joint box)
Steel thread
connection
(tool joint pin)
Aluminum
body pipe
drill string and drill pipes, fatigue failures on drill pipes
Steel pipe
Steel pipe
7/28/2019 2006_2p
8/35
8/35
Short introduction to drilling technology
Steel construction
fatigue locations
steel heavy construction vs. aluminum light construction
Aluminum construction
fatigue locations
Box fatigue
locationPin fatigue
location
Last Engaged Thread
- Notch effect
- Mean stress effect
(particularly for pin side)
Conical shoulder
Aluminum-Steel interface:
- Fretting nucleation
(different material stiffness)
steel
Fatigue location Fatigue location
Boxside
alluminum steel alluminum
Pinside
7/28/2019 2006_2p
9/35
9/35
Full scale fatigue testsdescription of test rigs
Bending arms SpecimenRotating masses Straingauge
1 m
Test rig forsteel construction
7/28/2019 2006_2p
10/35
10/35
description of test rigs
Test rig forsteel construction
F
t
F
t
-
Specimen
Rotating eccentric masses
--
Bending armBending arm
F2
Zt
de
H
The phase between the two couple of eccentric masses control the
stress amplitude
Full scale fatigue tests
7/28/2019 2006_2p
11/35
11/35
description of test rigs
Test rig forsteel constructionDevice to change the phase
Bending
arms
Supporting springs
to allow oscillatingdisplacements Specimen
Full scale fatigue tests
7/28/2019 2006_2p
12/35
12/35
description of test rigs
Test rig foraluminum construction
Full scale fatigue tests
Eccentric
rotating
mass
Rubber
wheels
Connection
to test
Electric
motor
Eccentric
rotating
mass
Stillmass Connection
to test
Rubberwheels
7/28/2019 2006_2p
13/35
13/35
description of test rigs
Test rig foraluminum construction
Aluminum pipe
Steel tool joint
Fatiguesection
Fatigue
sectionAluminum pipe
0.5 m
Steel tool joint
Strain gauge
Full scale fatigue tests
7/28/2019 2006_2p
14/35
14/35
description of test rigs
Test rig foraluminum construction
X
Y
Z
Deformed shape
Undeformed shape
Fix
point 2
Fix
point 1
Eccentric rotatingmass
Specimen prop
at fix points
Full scale fatigue tests
7/28/2019 2006_2p
15/35
15/35
description of test rigs
Full scale fatigue tests
ResonantTestRig.avi
7/28/2019 2006_2p
16/35
16/35
description of test rigs
The role ofresonance
Frequency
f, Hz
Bending stress amplitude
0 , MPa
Resonance
conditionIdeal
behavior true behavior(damping)
For different
masses or phases
Working frequency window, near the resonance condition.
High slope, good for control.
Full scale fatigue tests
7/28/2019 2006_2p
17/35
17/35
test results
Steel construction test results
Experimentalnucleation is fatigue life
when the smaller crack can be detected
through dynamic behavior control.
The Exp. Nucleation life includes a large
portion of propagation fatigue life.
In other words nucleation/propagation can
be resolved only when a large fatigue
crack appears in the structure.
Only pin side failure have obtained in this
fatigue test set10
5
106
107
108
0
20
40
60
80
100
120
cycles
0
[M
Pa]
Exp. nucleationFatigue lifeExp. nucleation fit lineFatigue life fit line
Full scale fatigue tests
7/28/2019 2006_2p
18/35
18/35
test results
Fatigue fracture section (pin)
fatigue crack starting from last
engaged thread root
High toughness leads to a large wall-through crack, before brittle fracture
(material:AISI 4145H)
Steel construction test results
Full scale fatigue tests
Crack fronts
2.5 cm
Detectable size
(exp. nucleation)
7/28/2019 2006_2p
19/35
19/35
test results
Aluminum construction test results
105
106
107
108
0
20
40
60
80
100
120
140
cycles
0
[MP
a]
TestsFit lineThe aluminum alloy here used
shows brittle behavior.
Then propagation phase can
not be distinguished from
dynamic behavior.
Full scale fatigue tests
7/28/2019 2006_2p
20/35
20/35
Full scale fatigue teststest results
Aluminum construction test results
Crack surface, showing:- initiation point
- brittle behavior
Fracture toughness is not enough
to allow wall-through crack.
(material AA 7014-T6)
After reaching this front, brittle fracture
happens.
Until this condition, dynamic behavior
control is almost steady.
2 cm
7/28/2019 2006_2p
21/35
21/35
Finite Element simulationsFE model dedicated to threaded connection
Steel construction FE model
Under bending load the stress state is
biaxial at the thread root surface:
r= 0
z> > 0
r = rz= z= 0
~ 0
Stress state is similar to plain strain condition.
r
z
Thread
root
Thread axis
direction
The make up produces a strong
presetting, and then a plastic zone
around the thread root can be found.
7/28/2019 2006_2p
22/35
22/35
Finite Element simulationsSteel construction FE model
Elastic shakedown at the last engaged thread root after
presetting:- linear kinematic hardeningcan be assumed
- limitedsubsequent stress amplitude
Subsequent
cycles
z
z
Presetting
z
m
za
z
p ~ 0
zp > 0
rp ~ -zp
1
1
e ~ 0
FE model dedicated to threaded connection
7/28/2019 2006_2p
23/35
23/35
Finite Element simulations
2D axial symmetry, to
avoid cumbersome 3D
analysis
Steel construction FE model
Element
discretization
at thread root
Bondedcontact
condition
Element
discretization
at thread root
Bondedcontact
condition
X
Y
Z
Axial
simmetry
Box
PinX
Y
Z
Axial
simmetry
Box
Pin
Elasto-
Plastic
material
model
Perfectelastic
material
model
Elasto-
Plastic
material
model
Perfectelastic
material
model
Elasto-plastic material
model, with linear kinematic
hardening behavior
Contact is modeled as
closed gap since no contact
loss is assumed.
FE model dedicated to threaded connection
7/28/2019 2006_2p
24/35
24/35
Finite Element simulationsSteel construction FE model
0 1 2 3 40
300
600
900
1200
1500
0 1 2 3 40
0.002
0.004
0.006
0.008
0.01
Stress path coordinate [mm]
S
tresses[MPa] pl
zr
Equivalentplastics
trainpl
Stress
path
Stress path along thread root bisector, afterpresetting
FE model dedicated to threaded connection
7/28/2019 2006_2p
25/35
25/35
Finite Element simulationsSteel construction FE model
Stress
path
Stress path along thread root bisector, afterelastic analysis.
0 1 2 3 40
300
600
900
1200
1500
Stress path coordinate [mm]
Stresses[M
Pa]
z/2/2r/2
FE model dedicated to threaded connection
7/28/2019 2006_2p
26/35
26/35
Finite Element simulationsSteel construction FE model
0 1 2 3 40
300
600
900
1200
1500
Stress path coordinate [mm]
Stresses[MPa]
z/2/2r/2
0 1 2 3 40
300
600
900
1200
1500
0 1 2 3 40
0.002
0.004
0.006
0.008
0.01
Stress path coordinate [mm]
Stresses[M
Pa] pl
zr
Equivalentplasti
cstrainpl
Subsequent
cycles
z
z
Make up plus
first cycle
z
m
za
FE model dedicated to threaded connection
7/28/2019 2006_2p
27/35
27/35
Fatigue modelsclassic approach (Gerber, kf, surface effect)
To propose a valid fatigue model the following issues need to be considered:
- reference S-N curve, with plain specimens, to relate load to fatigue finite life
- mean stress effect
(the strong presetting of the connection induce high tensile stresses)
- notch effect
(high gradient at the thread root)
- surface state effect(the machining to generate thread geometry can play a role in terms of fatigue nucleation)
Steel construction fatigue life prediction model
7/28/2019 2006_2p
28/35
28/35
Fatigue modelsreference S-N curve
Several plain specimen were extracted from real component to test as close as possiblein terms of:
- heat treatment,
- grain orientation.
Nf
a
classic approach (Gerber, kf, surface effect)
7/28/2019 2006_2p
29/35
29/35
Fatigue models
To take into account mean stress, the Gerber(parabola) model is considered.
Gerber parabola shows better fit with plain specimen extracted from real component
tested at positive mean stress ratios.
mean stress effect
a
m
classic approach (Gerber, kf, surface effect)
7/28/2019 2006_2p
30/35
30/35
Fatigue models
To take into account notch effect the following steps were considered:
- Same notch radius to determine the fatigue notch factorkf
- Also notched specimen are extracted from real component, and the notch bisector has same
orientation as thread root bisector
notch effect
classic approach (Gerber, kf, surface effect)
7/28/2019 2006_2p
31/35
31/35
Fatigue models
Finally particular care is dedicated to the surface effect:
- Small scale specimen extracted from thread geometry were tested to reproduce as close as
possible surface conditions
surface effect
-
classic approach (Gerber, kf, surface effect)
7/28/2019 2006_2p
32/35
32/35
Fatigue modelstest results correlationThe correlation is here presented as:
- Equivalent stresses against material limit at different cycles (left)
- logNpredicted logNExp.Nucl. diagram (right)
Wide discrepancy in terms
of cycles.
Not so bad in terms of stresses.
Eq. mean stress [MPa]
Eq.alternates
tress[MPa]
Failures
No failuresRun out
103 cycles
104
105
5 105
Fatigue limit 0
Pin stresses
Box stresses
00
100
200
200
300
400
400
500
600 800 100010
3
104
105
106
10710
3
104
105
106
107
Model prediction [cycles]
Exp.nucleation[cycles]
Tests
7/28/2019 2006_2p
33/35
33/35
Fatigue modelstest results correlationPossible sources of mismatch:
- bad assessment of mean stress (uncertainty of make up presetting, possible material cyclic
relaxation since it cycles at high mean stress)
- big portion of propagation
Eq. mean stress [MPa]
Eq.alternatestress
[MPa]
Failures
No failuresRun out
103 cycles
104
105
5 105
Fatigue limit 0
Pin stresses
Box stresses
0
0
100
200
200
300
400
400
500
600 800 1000
103
104
105
106
107
103
104
105
106
107
Model prediction [cycles]
E
xp.nucleation
[cycles]
Tests
7/28/2019 2006_2p
34/35
34/35
Conclusions Demanding full scale fatigue tests were proposed along with
the description of resonance test rigs.
Finite element dedicated to thread geometry was presented
- elastic-plastic analysis was needed for the high presetting,
- kinematic hardening was able to model elastic shakedown
Fatigue model proposed deals with simple tools for fatigueevaluation (Gerber, kf, surface effect) and calibration of the
model is based on small scale specimen as close as
possible to real component conditions.
To improve the correlation fatigue crack propagation should
be included, but:
how much is the nucleation/propagation crack length??
7/28/2019 2006_2p
35/35
35/35
ConclusionsWereexpensivefull scale fatigue tests necessary ??
YES, because:
- Some fatigue issues are hard to be thought a-priori.
- From small to full scale, propagation can play an important
role. Though prediction is conservative, large mismatch can be
found.
If we have toavoidfull scale testing:
Specimens, as close as possible to real component conditions,are needed, to calibrate fatigue models.