OTC-29223-MS

A Novel Approach to Fatigue Life Assessment of Subsea Connectors

Derek James Bennet, Andrew Macdonald Carmichael, and Simon John Roberts, Expro North Sea Ltd

Copyright 2019, Offshore Technology Conference

This paper was prepared for presentation at the Offshore Technology Conference held in Houston, Texas, USA, 6 – 9 May 2019.

This paper was selected for presentation by an OTC program committee following review of information contained in an abstract submitted by the author(s). Contents of
the paper have not been reviewed by the Offshore Technology Conference and are subject to correction by the author(s). The material does not necessarily reflect any
position of the Offshore Technology Conference, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Offshore Technology Conference is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of OTC copyright.

Abstract
The capability to predict fatigue damage continues to be critical for determining the operational life and
inspection intervals of connectors and components used in offshore applications. Subsea well intervention
systems are subjected to wave induced cyclic bending moments and understanding the fatigue performance
of equipment is essential for determining safe operating envelopes.
In this paper, a validated fatigue analysis methodology is presented for non-preloaded connectors that
are used within subsea well intervention systems. The fatigue analysis methodology addresses limitations
in current standards when calculating the fatigue capacities of non-preloaded connectors with different
interacting component materials (i.e. low alloy steel and nickel based alloys). The methodology considers
the effect on the fatigue life of both non-axisymmetric geometry/loading, FAT loading, as well as the
interaction of different connector materials, capturing any potential change in hot spot locations.
Three different non-preloaded connections (i.e. consisting of threaded and load shoulder connectors)
were analysed using 3-D finite element analysis models, where ΔM-N curves and the associated crack
initiation locations were calculated for each connector. Full-scale resonance fatigue tests were successfully
performed on the three different connector types, validating the ΔM-N curves calculated using the fatigue
analysis methodology. Fatigue failure (i.e. through-wall crack) was achieved in all tests between 100,000
and 5,000,000 cycles matching the predicted crack initiation location from the analysis for each connection.
The validated methodology provides accurate calculation of the fatigue life and correct identification of hot
spot locations. Using the validated approach described in this paper within the design process can lead to
significant improvements in future designs of connectors, enabling safer operational limits and extending
the service life of subsea systems.

Introduction
Connectors within the subsea well intervention system are defined as a mechanical or hydraulically activated
mechanism that is used to connect components together, transfer load and prevent leakage [ISO 13628-7.
2006]. Examples of connectors include, but are not limited to, threads and load shoulders. Ensuring the
safe design of these connectors is critical for determining the operational life and inspection intervals of
connectors and components used in offshore applications [Chen et al. 2015]. In particular, protection against
fatigue failure is essential as connectors within these systems are often exposed to significant cyclic loading
2 OTC-29223-MS

due to wave induced motion or from operational effects such as the movement of the rig [Kirkemo 2014].
When subjected to these cyclic loads, geometrical stress raisers can initiate fatigue cracks that cause the
system to fail, resulting in a shortened lifetime [Claeys et al. 2012].
Governing standards for the assessment of fatigue load capacity of subsea connectors require that both
analysis and testing are performed [ISO 13628-7. 2006, API RP 17G. 2006, DNVGL-RP-C203. 2016]. A
best practice guide for establishing the design fatigue load capacity of subsea well intervention connectors
is provided in Worsen et al. [Worsen et al. 2017] and contains two methods for preloaded and non-
preloaded connectors: either by analysis only (preloaded only) or by analysis and full-scale fatigue testing
(preloaded or non-preloaded). If full-scale fatigue testing is performed, then in order to validate the analysis
methodology, a minimum of one full-scale fatigue test to failure is required. Through the use of the test
curve target approach, the analysis methodology is validated if the tested mean fatigue life is greater than
the fatigue test curve [BS 7608. 2014, Johnston 2017].
To determine the fatigue load capacity of connectors, a common approach is to use finite element analysis
(FEA) methods [Cetin et al. 2010, Lin et al. 2016, Schneider et al. 2010, Perales et al. 2017] with suitable
S-N fatigue curves [Kirkemo 2014 and Wormsen et al. 2017]. Design S-N curves provided in DNVGL-
RP-C203 [DNVGL-RP-C203. 2016] can be used for components made from high strength steels that are
applicable to a high mean tensile stress (HS curve) or a limited range of mean stress (BM curves). In order
to more accurately predict the fatigue load capacity and failure location of the non-preloaded connectors
considered in this paper, material testing was performed to develop material S-N curves at zero mean stress.
The fatigue load capacity should be presented in terms of cyclic bending moment range versus number of
cycles to failure. This approach is aimed at standardising and simplifying the interface between the local
connector analysis and the global riser analysis of the full subsea well intervention system. The effect on
the fatigue life of local yielding due to pressure test loading as well as both non-axisymmetric geometry
and loading should be considered in the calculation of the fatigue load capacity.
To perform full-scale fatigue testing there are several different approaches that may be used for example
four-point bending or resonance bending. Resonance bending is often used as the test specimen can be
cycled to the desired bending moment at much higher frequencies compared with the four-point method
resulting in significantly reduced test time [Wittenberght et al. 2012].
In this paper, three different non-preloaded connections, defined as a collective term to describe an
assembly of components consisting of at least one connector, were analysed using 3-D FEA. The fatigue
analysis methodology addresses limitations in current standards when calculating the fatigue capacities of
non-preloaded connectors with different interacting component materials (i.e. low alloy steel and nickel
based alloys). It also highlights the importance of accurate modelling of non-axisymmetric features as well
as the residual stress fields due to FAT loading prior to cyclic loading. Full-scale resonance fatigue tests
were then performed on each connector type in order to validate the predicted fatigue load capacity. Fatigue
failure (i.e. through-wall crack) was achieved in all tests between 100,000 and 5,000,000 cycles matching
the predicted crack initiation location from the analysis for each connection. The test mean fatigue life was
greater than the fatigue test curve for all connections, validating the analysis methodology.

Non-preloaded Connections
The connector(s) and materials associated with the three non-preloaded connections considered in this paper
are summarised in Table 1 with representative geometries shown in Figure 1.
OTC-29223-MS 3

Table 1—Summary of Connections

Connection Material
Connector
ID Male Component Female Component

1 Thread type 1 Low alloy Steel Low alloy Steel

2 Thread type 2 Low alloy Steel Nickel based alloy

3 Thread type 2 & load shoulder Low alloy Steel Nickel based alloy

Figure 1—Connection 1, 2 and 3 geometry types

Fatigue Analysis
Finite Element Modelling
A 3-D half model was developed for each connection in Table 1 using ANSYS Mechanical APDL version
17.0. The model geometry considered all the fatigue limiting features based on the worst case manufacturing
tolerances. Non-axisymmetric features considered critical to the fatigue load capacity, such as radial holes
and slots, were also included in the model. An example of the model geometry for Connection 2 is shown
in Figure 2.

Figure 2—Connection 2 model geometry

The fatigue analysis material model used an elastic-plastic cyclic stress-strain curve with kinematic
hardening to capture local residual stress at hot spots due to pressure loading. Low order 6-sided hexahedral
elements were utilised in order to attain a solution within a reasonable timeframe. Degenerate elements were
used at critical regions of complex geometry with justification.
Initially the 3-D models were meshed to capture the global response of the connection. Once the hot
spots were identified, a 3-D submodel was then generated from the global model to attain a mesh that could
produce sufficiently accurate results in those regions. Mesh sensitivity studies were performed to ensure the
4 OTC-29223-MS

mesh was suitably refined to capture local stresses. A plot of the global and submodel meshes for Connection
2 are shown in Figure 3.

Figure 3—Connection 2 global and submodel mesh

The models were constrained along the longitudinal cut-boundary and a cross-sectional cut-boundary
using symmetry boundary conditions. Tensile and bending loads were applied to the unconstrained cross-
sectional cut-boundary. Bore pressure was applied to all internal wetted surfaces. All interfaces between
components were modelled with a representative coefficient of friction considering surface finish, any
lubrication, contact pressure, etc.

S-N Curve Selection

Material testing was performed to develop material S-N curves at zero mean stress, in accordance with the
procedures provided in ISO 1143 [ISO 1143. 2010] and statistical analysis provided in ISO 12107 [ISO
12107. 2012].

Stress Range Calculation

The 3-D global model was subjected to the loading stages shown in Figure 4. It is noted that although
an elastic-plastic material model was used, there was no plasticity in the cyclic load stage. Therefore, the
stresses used to calculate the stress range and mean stress resulted from an elastic load and unload cycle.

Figure 4—Fatigue analysis loading stages

OTC-29223-MS 5

To calculate the stress range, the Sines (von Mises based) criterion was used [DNVGL-RP-C203. 2016].
The von Mises stress ranges between load condition t and t’ at each fatigue hot spot was calculated using
Equation 1:

(1)

To calculate the mean stress, the Rudolph and Weiß criterion was used which is a modified version of
the Sines mean stress [Worsen et al. 2015]. The mean stress between load condition t and t’ was calculated
using Equation 2:

(2)

The calculated stress range was then corrected to take account for the effects of mean stress using the
Morrow relationship [Morrow 1968].

Calculation of ΔM-N Curves

Strength reduction factors are critical to predict the translation of the fatigue life data of small scale, rotating
bend material samples to large scale, and factory finished components under actual loading. These factors
provide reduction to the stress range of the tested mean material S-N curve at a specific number of cycles
to failure. The Juvinall method of reducing the material S-N curves to component S-N curves was applied
[Juvinall 2012].
For each of the applied loading ranges, the maximum mean corrected stress range and the component S-
N curve were used to calculate the number of cycles to failure and mean ΔM-N curve for each hot spot.
A lower bound mean ΔM-N curve was constructed from the mean ΔM-N curves for each fatigue hot spot.
Target ΔM-N curves were then calculated for testing purposes ensuring that the test results would have
sufficient conservatism with the lower bound mean ΔM-N curve calculated from the FEA.

Fatigue Testing
Full-scale resonance fatigue testing of each connection should target either low, medium or high fatigue
loads defined as follows [Wormsen et al. 2017]:

• High load with target fatigue life of 100,000 – 500,000 cycles

• Medium load with target life of 500,000 – 2,000,000 cycles

• Low load with target life of 1,000,000 cycles – 5,000,000 cycles.

The fatigue analysis methodology presented is validated if the connections tested have a mean fatigue
life greater than the fatigue test curve. The test fatigue failure location should match the calculated fatigue
failure location.

Resonance Fatigue Testing Theory

In resonance fatigue testing, a test specimen is subjected to an eccentric rotating mass, exciting the test
specimen close to its natural frequency. This results in a bending moment applied to the test specimen, such
that the full circumference is exposed to the same bending moment during one cycle. When the test specimen
vibrates in the first mode, the maximum bending moment is found at the middle of the test specimen, with
zero deflection occurring at two positions corresponding to the first eigenmode.
Tests are run below the resonant frequency, such that the deflection and strain response of the specimen
can be controlled. Once a desired bending moment is known, strain gauges on the pipe sections can be used
6 OTC-29223-MS

as an input into the test. The frequency of the excitation mass can then be increased until a desired axial
strain range is achieved in the pipe, corresponding to a desired bending moment in the region of interest.
The resonance bend test is then left running until either fatigue failure occurs or a defined number of cycles
is reached indicating that the stresses from the selected bending moment are below the endurance limit of
the specimen.

Resonance Fatigue Test Set-up

A simplified setup of the test rig and test specimen is shown in Figure 5. The test specimen consists of
two stabilising masses, with pipes either side of the connection. The test specimen is supported at two
locations, corresponding to the calculated zero deflection positions of the test specimen. A motor rotates
the eccentric mass resulting in bending moment applied to the test specimen. The test specimen, including
the connection design, was such that the resonant frequency of the specimen could be achieved with the test
rig. All tests were performed in air, at room temperature. As the connections considered in this paper are all
non-preloaded, preliminary testing was performed on each connection to determine the minimum internal
pressure required to achieve stable resonance over the desired strain range.

Figure 5—Resonance fatigue test set-up

To provide an input into test and also to monitor the response of the test specimens, strain gauges were
applied and monitored during test. The frequency of the eccentric mass was increased until a desired average
axial strain range in the pipe was achieved, correlating to a desired bending moment at the connection. To
detect if fatigue failure had occurred a pressure transducer trip was implemented to detect loss of internal
pressure and a strain gauge trip was used on the connector to detect a change in stiffness.

Harmonic Analysis
The strain data calculated from static structural analysis based on the ΔM-N target test curve with a
constant bending moment does not provide an accurate input for resonance bending testing, where the
load distribution will be nonlinear. In order to provide accurate strain data at the strain gauge locations for
the resonance fatigue testing, harmonic FEA of the test specimen was performed. Harmonic analysis is a
technique to determine the steady state response of a structure to sinusoidal loads at a known frequency.
The harmonic analysis represented the test specimen and resonance fatigue test set-up shown Figure 5.
The test specimens were modeled as a continuous straight pipe with varying cross-sections using 3-D beam
elements. Stablising mass objects located at the ends of the test specimen were modeled as a point mass.
To represent the support from the test rig, the model was constrained as a simply supported beam. Prior
to performing the harmonic analysis, a static structural analysis was performed to pre-stress the model and
represent the stiffness of the test specimen when subjected to bore pressure and associated end cap effect.
The dynamic loading of the excitation mass was applied to the pre-stressed model to determine the steady
state response of the test specimen. Figure 6 shows the steady state response of the test specimen for a given
frequency in terms of axial strain.
The results of the harmonic analysis provided a relationship between the axial strain range at a strain
gauge location on the plain pipe section and the bending moment range at the connector. The relationship
OTC-29223-MS 7

was used to provide strain range data at strain gauge locations to subject the test specimen to the required
cyclic bending moment ranges. The calculated axial strain distribution was also used as a form of verification
between analysis and test results.

Figure 6—Harmonic analysis – steady state response

Results and Discussion

Prior to performing the analysis, preliminary testing was performed on each connection to determine the
minimum internal pressure required to achieve stable resonance over the desired test axial strain range. The
results of the global FEA model were used to identify fatigue hot spots for the applied cyclic loading, as
shown in the example provided in Figure 7. Thereafter, a submodel was created of the identified hot spot
locations to provide more accurate stress results as shown in the example provided in Figure 8.

Figure 7—Regions of high stress range (grey contours are regions of low stress range)
8 OTC-29223-MS

Figure 8—Submodel of hot spot location

Figure 9 shows mean ΔM-N curves, calculated from FEA for multiple hot spots. The curves demonstrate
the importance of considering all hot spots when calculating the connection mean ΔM-N curve, which uses
the lower bound of the mean ΔM-N curves from all the hot spots. It is also shown in Figure 9 the potential
for the hot spot to change at different bending moment ranges when connections consist of components
with different materials.

Figure 9—Example of mean ΔM-N curves for each identified hot spot

Fatigue tests were then performed on each connection, targeting either low, medium or high fatigue
loads. In all tests, once the target average axial strain was achieved in the pipe, the strain gauge response
was inspected to confirm that the specimen was responding as expected. An example of the axial strain
gauge response against results from the harmonic model is provided in Figure 10 showing good correlation
between test and analysis.
OTC-29223-MS 9

Figure 10—Comparison between analysis and test axial strain gauge response

To verify that all tests were resonating in the same mode, the strain gauge response at each strain gauge
set was normalised against the target average axial strain in the pipe. As can be seen in the example provided
in Figure 11, this verified that the all tests were resonating in a similar way.

Figure 11—Normalised average axial strain response against target strain gauge set

Fatigue failure (i.e. a through-wall crack) was achieved in all tests between 100,000 and 5,000,000 cycles
matching the predicted crack initiation location from the analysis of each connector. This was identified by
a pressure trip in Connection 1 and by a strain gauge trip in connections 2 and 3. For connections 2 and 3,
once the test had stopped, visual inspection was performed. If a crack could not be found on the surface of
the test specimen, the test was continued in order to propagate the crack further. An example of the strain
gauge response is shown in Figure 12. As shown, stable resonance was initially achieved and then a strain
gauge trip caused the test to stop. As no crack was identified on the surface of the test specimen, the test was
restarted twice more in order to propagate the crack and achieve a through-wall crack. Once a through-wall
crack was identified, as shown in the example provided in Figure 13, the test specimen was drained and
disassembled. NDE inspection confirmed the crack initiation location on the ID of the component inspected.
10 OTC-29223-MS

To further inspect the crack initiation location, the crack was broken open to inspect the fracture face, as
shown in Figure 13. As can be seen, beachmarks were found propagating from the crack initiation location.

Figure 12—Example of strain gauge response

Figure 13—Example of through-wall crack identified on OD of component and crack broken open

In all connections the mean test fatigue life was above the target fatigue curve, validating the analysis
methodology, as shown in the figures 14-16.
OTC-29223-MS 11

Figure 14—Calculated ΔM-N curves and test results for Connection 1

Figure 15—Calculated ΔM-N curves and test results for Connection 2

12 OTC-29223-MS

Figure 16—Calculated ΔM-N curves and test results for Connection 3

Conclusions
In this paper, a validated fatigue analysis methodology has been presented for calculating the fatigue load
capacity of non-preloaded connectors. The fatigue analysis methodology addresses limitations in current
standards when calculating the fatigue capacities of non-preloaded connectors with different interacting
component materials (i.e. low alloy steel and nickel based alloys), capturing any potential change in hot
spot locations. The analysis presented in this paper also highlights the importance of accurate modelling
of non-axisymmetric features as well as incorporating residual stress fields due to FAT loading prior to
cyclic loading. Three different non-preloaded connector types have been analysed and tested, validating the
analysis methodology which provided accurate calculation of the fatigue life and correct identification of
hot spot locations. Using the validated approach described in this paper within the design process can lead
to significant improvements in future designs of connectors, enabling safer operational limits and extending
the service life of subsea systems.

Acknowledgements
We would like to thank Weihang Chen for his contribution to the development of the analysis methodology
and The Welding Institute (TWI) for their support with the fatigue testing.

Nomenclature
∆σeqν von Mises stress range between time instances t and t’.
∆σx x-component of stress range between time instances t and t’.
∆σy y-component of stress range between time instances t and t’.
∆σz z-component of stress range between time instances t and t’.
∆σxy shear stress range in the xy-plane between time instances t and t’.
∆σyz shear stress range in the yz-plane between time instances t and t’.
∆σxz shear stress range in the xz-plane between time instances t and t’.
OTC-29223-MS 13

von Mises mean stress between time instances t and t’.

x-component of mean stress between time instances t and t’.
y-component of mean stress between time instances t and t’.
z-component of mean stress between time instances t and t’.
shear mean stress in the xy-plane between time instances t and t’.
shear mean stress in the yz-plane between time instances t and t’.
shear mean stress in the xz-plane between time instances t and t’.

References
API RP 17G. Recommended Practice for Completion/Workover Risers. Identical to ISO 13628-7, 2006.
BS 7608. Guide for fatigue design and assessment of steel products. 2014.
Cetin, A. Härkegård, G. Fatigue life prediction for large threaded components. Procedia Engineering 2 (2010) 1225 –
1233.
Chen, W., Orlando, M., Roberts, S., Design Methodology for the Structural Capacity and Shakedown Assessment of Bolted
Joint Connections in Subsea Open Water Systems. SPE-175444-MS. SPE Offshore Europe Conference & Exhibition
held in Aberdeen, United Kingdom, 8–11 September 2015.
Claeys, J., Wittenberghe, J.V., Baets, P.D., Waele, W.D. Characterisation of a resonant bending fatigue test setup for pipes.
Sustainable Construction and Design. 2011.
DNVGL-RP-C203. Fatigue Design of Offshore Steel Structures. Recommended Practice. DNV, 2016.
ISO 13628-7. Petroleum and natural gas industries – Design and operation of subsea production systems. Part 7:
Completion/workover riser systems. ISO, 2006.
ISO 1143. Metallic materials - Rotating bar bending fatigue testing. 2010.
ISO 12107. Metallic Materials - Fatigue Testing - Statistical Planning and Analysis of Data. 2012.
Johnston, C. Statistical Analysis of Fatigue Test Data. Proceedings of the ASME 2017 36th International Conference on
Ocean, Offshore and Artic Engineering, June 25 – 30, 2017, Trondheim, Norway.
Juvinall, R.C., Marshek, K.M. Machine Component Design, Fifth Edition. Singapore: John Wiley & Sons. 2012.
Kirkemo, F. Structural Design of Metallic Components in ISO 13628-7. Proceedings of the ASME 2014 Pressure Vessels
& Piping Conference. July 20 – 24, 2014. Anaheim, California, USA.
Lin, T., Zhang, Q., Lian, Z., Liu, Y., Zhang, Y., Chen, Y. Multi-axial fatigue life prediction of drill collar thread in gas
drilling. Engineering Fatigue Analysis 59 (2016) 151 – 160.
Morrow, J. SAE Fatigue Design Handbook, 1st edition SAE AE-4 Society of Automotive Engineers Inc 1968.
Schneider, R., Wuttke, C., Berger, C. Fatigue analysis of threaded connections using the local strain approach. Procedia
Engineering 2 (2010) 2357 – 2366.
Perales, S.A., Kaculi, J.T. Verification Analysis and Validation Testing of Subsea Connectors. Proceedings of the Offshore
Technology Conference, May 1 – 4, 2017, Houston, Texas, USA.
Wittenberghe, J. V., Baets, P. D., Waele, W. D., Ost. W., Verstraete, M., Hertele, S. Resonant Bending Fatigue Test Setup for
Pipes with Optical Displacement Measuring System. Journal of Offshore Mechanics and Artic Engineering. ASME,
2012, Vol. 134 / 031702-1.
Wormsen, A., Kirkemo, F., Muff, A.D. Best Practice Guidance for Establishing the Design Fatigue Capacity of Subsea
Well Intervention System Connectors. Proceedings of the ASME 2017 36th International Conference on Ocean,
Offshore and Artic Engineering, June 25 – 30, 2017, Trondheim, Norway.
Wormsen, A., Avice, M., Fjeldstd, A., Reinas, L., Macdonald, K.A., Berg, E., Muff, A.D. Fatigue testing and analysis of
notched specimens with typical subsea design features. International Journal of Fatigue 81 (2015) 275 – 298.