OTC-25779-MS Chain Out of Plane Bending (OPB) Joint Industry Project (JIP) Summary and Main Results
OTC-25779-MS Chain Out of Plane Bending (OPB) Joint Industry Project (JIP) Summary and Main Results
OTC-25779-MS Chain Out of Plane Bending (OPB) Joint Industry Project (JIP) Summary and Main Results
Chain Out of Plane Bending (OPB) Joint Industry Project (JIP) Summary and
Main Results
Lucile Rampi, SBM Offshore
Fata Dewi, SBM Offshore
Pedro Vargas, Chevron
This paper was prepared for presentation at the Offshore Technology Conference held in Houston, Texas, USA, 4–7 May 2015.
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
In 2002, several mooring chains of a deepwater offloading buoy failed prematurely within a very small time frame. These
chains were designed according to conventional offshore fatigue assessment using API recommendations. With this first
deepwater buoy application, a new mooring chain fatigue mechanism was discovered. High pretension levels combined with
significant mooring chain motions caused interlink rotations that generated significant Out of Plane Bending (OPB) fatigue
loading. Traditionally, interlink rotations are relatively harmless and generates low bending stresses in the chain links. The
intimate mating contact that occurs during the proof loading and the high pretension of the more contemporary mooring
designs have been identified as aggravating factors for this phenomenon.
A Joint Industry Project (JIP), gathering 28 different companies, was started in 2007 to better understand the OPB mooring
chain fatigue mechanism and propose some mooring chain fatigue design recommendations.
This paper summarizes the various test programs that were implemented within the more than 6 years long project, including
full scale fatigue tests on chains, a quasi static OPB stiffness measurement campaign, and tests on small samples addressing
the environmental parameter influence on fatigue initiation and crack propagation stages.
The main output from the FEA scope of work, performed to support the experimental tests, will also be described.
Finally, the paper will address the major step that has been achieved regarding implementation of a standard practice in
offshore industry using a multiaxial fatigue criterion to address OPB hotspots.
Introduction
Out of Plane Bending (OPB) of a chain link refers here after to the bending of a chain link out of its “main plane” (the plane
containing the oval shape, see Figures 1 and 4). It is caused by the application of transverse forces and OPB moments which
are resisted by frictional forces at the contact between links.
In Plane Bending of a chain link (refers hereafter as IPB) is related to the bending of a chain link inside its “main plane”. This
type of loading has also been considered within the frame of the project. This loading is however less stringent for fatigue
because the nominal stress related to IPB is approximately seven times smaller than the nominal stress related to OPB loading
(in line with both bending inertia ratio). Both OPB and IPB fatigue loadings are experienced in addition to the conventional
chain tension fatigue mechanism.
2 OTC-25779-MS
Chainhawse
Figure 1 : Chain Link Failure on deepwater buoy due
to Out of Plane Bending (OPB) Figure 2 : Deepwater buoy and chain connection to the
floating body
Straight Section
Bend Section
Crown Section
Under high pretension, the interlink friction is high and the moment to be resisted at the contact between links can be
significant. This rather unusual loading condition for a chain link arises from the combination of high pretension load, as a
percentage of Minimum Breaking Load (MBL), and poor articulation of the fairlead or chainhawse (chain connecting system
with the floating body as shown in Figure 2). The alternating OPB stresses may be large enough to cause the fatigue crack
propagation in the first moving link adjacent to the chain connection at fairlead and failure can occur rapidly. On particular
case of the first deepwater buoy (see Figure 2), 4 links have failed in less than 8 months always at the same location: on the
first moving link.
Typical deepwater mooring pretension can be as high as 15% of the MBL of the chain. As an example, for a 127 mm grade R4
studless chain, this is about 230 tonnes.
A key difference between the traditional mooring chain tension fatigue mechanism and OPB fatigue mechanism is that chain
tension fatigue uses the stress histogram applied to the mooring chain neglecting chain rotation effects, whereas OPB fatigue
uses the inter-link angular rotation converted to a stress (under constant chain tension) as a source of fatigue. Figure 1 shows
the failure of a 81mm mooring chain in OPB. Figure 3 shows the different fatigue crack locations for tension and OPB fatigue
OTC-25779-MS 3
mechanisms.
OPB fatigue mechanism is therefore identified as the root cause of the short service life of the first deepwater buoy mooring
chains (see Ref [5] ).
After this event, SBM undertook several different test campaigns to reproduce the OPB mechanism (see Ref [6] Ref [7] and
Ref [8] ) and generate the first laboratory demonstration of the OPB fatigue.
In 2007, a Joint Industry Project (JIP) was proposed to the offshore industry with the main objective of implementing a full
scale chain testing program to better assess the OPB fatigue mechanism.
The detailed objectives of this JIP were to:
• Perform a quasi static test campaign on chains to assess the interlink contact stiffness
• Implement various FEA to assess the stress state at different locations and run sensitivity studies
• Implement a fatigue test program to get OPB endurance data points (S-N curve)
• Provide some design guidance to address the OPB chain fatigue mechanism
In addition, a test program on small samples was implemented within the frame of this JIP to better assess the influence of
environmental parameters on fatigue test (in air or in seawater testing) and its relation to the testing frequency by looking
separately at fatigue crack initiation and fatigue crack propagation stages.
All these testing program are summarized in Ref [1] .
Symbols definition
The objective of this campaign was to define OPB stress and its dependency upon the following parameters:
• Chain diameter effect: 4 chain size have been tested (84 mm, 107mm, 127mm, 146 mm)
• Tension effect: 5 level of static tension have been applied (5% MBL → 25% MBL)
• Grade effect: R3, R4 chain grades have been tested
• Manufacturing effect: Vicinay and Ramnas manufactured chains have been tested
• Shape effect: studless and studlink chains have been tested
• Environment effect: tests in air and in seawater have been performed
• Proof load effect: 1 non proof loaded sample has been tested
As illustrated in the following figure, the ‘static’ test consists of 15 links installed in a specific test rig with the following
conditions:
• constant static tension applied horizontally
• imposed vertical displacement on the sample middle link (link number 8)
• symmetrical up and down motions
• free rotation allowed at each extremity
Stroke
The test principle was to apply a static tension at one end and apply a vertical displacement on the center link which can slide
within a loading box. The center loading box moves up and down, generating some OPB moment on both adjacent links.
All static tests on chain samples were performed at Stress Engineering Services (SES’s) testing facility in Houston.
Figure 6 shows the test rig that was used to test smaller chains (84mm & 107 mm chains).
This bench is an existing 1 million pounds (LBS) frame modified to implement a vertical cylinder which displaces the center
link (link L8) up and down. The center link is placed in a center loading box (see Figure 7) which prevents link rotation and
allows some link sliding to maintain constant tension. Free rotations (rolling mechanism) at the extremities are achieved by
using special E-links and horizontal pins. The bench length limits the maximum sample length to approximately 25 feet, hence
it is only used for ‘static’ test of 84 mm and 107 mm diameter chain samples.
Another test bench was constructed for testing larger chains and for the full scale fatigue tests. It was based on a similar
concept as the first bench but it has 2 independent testing bays and it can accommodate 15 links of larger chain diameter. It
was also fitted with a water-trough system (for testing in seawater) and was designed to withstand the fatigue loading of all
fatigue samples.
OTC-25779-MS 5
Figure 7: Center loading box on small bench Figure 8: Dual bays fatigue test bench
Fourteen “static” tests were performed at very low cyclic frequency (20 cycles recorded for each static tension value).
One of these static tests was performed in artificial seawater.
The chain samples were instrumented to derive the stress value at given locations and to assess the angle distribution along the
length of the chain.
The following parameters were monitored and recorded during testing:
• The horizontal stroke and force (the horizontal tension shall remain constant)
• The vertical cylinder stroke and force
• The principal stresses values and directions at given locations
• The angle variation for the different links
The minimum instrumentation in place was to have links L6 to L10 instrumented with strain gauges (4 gauges per link) in the
link center part (see figure below) and inclinometers. Some samples were instrumented with more gauges, especially close to
the interlink contact area to assess some Stress Concentration Factors (SCF) and also with transductors displacement on all 15
links to check the angle variation distribution along the chain sample.
MOPB_straight part
MOPB_contact area
Gauges
Link 8
Figure 9: Strain gauge location Figure 10: 107 mm chain sample strain gauging
The measured principal stresses and angle variations data acquired from the tests are post-processed to correlate local interlink
stiffness K to OPB moment and interlink angle;
Equation 1: M OPB = K (T, d, α int , Proof load ) × α int .
The first step consisted in deriving the OPB bending stress in the link straight part function of the interlink angle. This was
done by combining the data from 4 gauges located in middle of the link straight part to get a representative average OPB
bending moment. For the IPB links, an equivalent OPB bending stress was derived from the calculated IPB moment using the
nominal OPB bending inertia (pseudo OPB stresses for IPB links). This enabled the stresses (or bending moment) coming
from OPB links (L7 & L9) to be directly compared to the one coming from the IPB links (L6 & L10).
6 OTC-25779-MS
Figure 11 below shows a typical hysteresis of the OPB stresses (pseudo OPB stresses for IPB links) in the middle of the link
straight part function of the interlink angle for OPB and IPB links for all recorded cycles.
Locking point
Figure 11: step 1 -Typical hysteresis - OPB bending Figure 12: Average curve for link 7 –up & down cycles
stresses function of the interlink angle
Then the next step (step 2 as shown in Figure 12 and Figure 13) was reducing all selected recorded cycles into one single
average cycle and then extracting the “going up” and “going down” part of the curve that were put in the same diagram to
compare the slopes. It can be highlighted that the OPB bending stress increases with the static tension applied during testing as
shown in Figure 13.
T=500 kips
T=400 kips
T=300 kips
T=200 kips
T=100 kips
Polynomial functions were then applied on each ‘going up’ and ‘going down’ OPB stress range curves functions of the
interlink angle curves (as shown on Figure 14). Recalculated stresses were then entered into one comprehensive resulting test
matrix containing all test related information; interlink angle range, OPB stress range, recorded static tension applied, chain
diameter, sample identification number, chain grade (R3 or R4), link number (L6, L7, L9, L10), cycle part (up or down) and
reference static tension value as shown in the extract table below (see Figure 15).
It has been demonstrated that a power law of the tension and diameter was appropriate to represent the bending moment or
bending stress at the connecting point of a chain subject to an inclined tension. The OPB stress range can be defined with the
following equation.
OTC-25779-MS 7
The figure hereafter shows the recalculated best fit surface together with test data results for given chain diameter. It provides
the OPB stress range (in the middle of the link straight part) function of the interlink angle range and the static tension applied.
∆σOPB d=107
mm
Strain gauge
location
Interlink
contact area
∆αint T
Figure 16: Recalculated OPB stress & tests data points
The OPB stress is derived from strain gauges placed on the link straight part. For the tested chain configuration, the bending
moment increases along the OPB links (L7 or L9) between the link straight part (where the uniaxial gauges are placed) and the
contact area, as shown in Figure 17. To derive the interlink contact stiffness, the bending moment at the interlink contact area,
must be defined. Therefore a bending moment transfer function was derived between the link straight part (gauge location) and
the interlink contact area. This was done by using finite elements analysis results that enabled derivation of the OPB stress as a
polynomial function of the static tension applied and chain diameter.
Regarding the interlink contact, different locking modes can be distinguished during a typical hysteresis cycle as shown in
Figure 11:
• Sticking locking mode: steepest part of the stress slope,
• Intermediate rolling mode: the curves are distinctively more nonlinear
• Sliding mode: the stress remains more or less constant
The hysteresis on both OPB links (L7 & L9) as discernible in Figure 11 indicates that interlink sliding threshold was reached.
The OPB moment will be limited on each link when the interlink sliding threshold is reached as defined by the following
equation:
8 OTC-25779-MS
d
Equation 5 ∆M OPB _ sliding = 2 * µ * T * = µ *T * d
2
Note that this sliding threshold definition is a simplified assumption based on friction slippage on nominal link diameter.
However it correlates well with the tests results.
Among all 14 static tests run, 13 tests were performed in air and one test was performed in seawater conditions. The direct
comparison between the test in air and the test in seawater using the same sample has clearly demonstrated that:
• The slopes at origin (sticking locking mode) are quite consistent for both environment conditions.
• Sliding occurs quicker in seawater than in air as illustrated in both following figures.
However, it has been clarified, thanks to the correlation with FEA, that the main source of scatter was linked to the variability
of the contact point location. The contact stiffness will be different if the contact point occurs inside or outside the intimate
contact area generated during chain proof load. This means that this scatter will exist for real chain in service. By testing a
large amount of chain links (2 locking points for the going up and going down curves, 4 different chain links monitored, 14
different static tests run, 5 different static tension value each time), the results can be considered as statistically representative .
The influence from various parameters tested during this static test campaign was also assessed by making direct comparison
between tests with parametric variations:
• Chain grade effect: comparison between grades R3 and R4
• Manufacturing effect: comparison between Vicinay and Ramnas samples
• Link shape effect: comparison between Studlink and studless shapes
At the end the variability on those parameters was not statistically representative and therefore they were not treated
separately.
OPB and IPB links have also been considered separately and they have shown a good consistency on the linear part of the
curve. The range of the interlink angle variation for IPB links was however significantly smaller than for OPB links.
Consequently only the results derived on OPB links were used to derive the interlink contact stiffness to reduce the scatter.
The final end result was one single model proposed to derive the interlink contact stiffness function of the interlink angle, the
tension applied and the chain diameter based on a statistically robust large set of results.
OTC-25779-MS 9
Several FEA scopes of work were performed by Principia during the course of this project to assess different aspects
summarized as follow:
• OPB locking mechanism understanding
• Sensitivity studies to various parameters
• Model the bench configuration
• Assessment of stress state at various locations
A large amount of non-linear elastic-plastic calculations were performed using 3D type of elements and Abaqus as a solver to
represent a few links subjected to, first, application and release of the tensile proof load value (a minimum of 70% of the chain
MBL) and, second, combined tension and bending.
Initial studies were performed to derive correlation between stresses in the link straight part to stresses in the contact area and
to derive an OPB stiffness law for interlink locking mode (SCFs, interlink contact stiffness, sensitivities to various
parameters). However, as more insights of chain OPB behaviour were gained, the studies have progressed into simulation of
chain testing in the test bench and assessment of multiaxial stress state in the link contact area.
Figure 20: Typical OPB link (link 2) FE model Figure 21: OPB link FEA with test bench constraints
For the interlink locking mechanism, the bending moment curve function of the interlink angle was considered as well as the
relative displacement for adjacent nodes as shown in Figure 22 and Figure 23.
This enables to distinguish the 3 different locking modes previously detailed when describing the static tests results (sticking,
rolling and sliding). It demonstrated that the stiffness curve non linearity significantly increases as soon as both adjacent nodes
start to move apart (intermediate rolling mode).
MOPB
Node2
Node1
MOPB N.m
The cyclic behaviour was assessed by performing full cycles analyses. The hysteresis obtained in static testing was
successfully simulated with FEA. It was also successfully demonstrated that different locking stiffness can take place
10 OTC-25779-MS
depending if the locking is inside or outside of the intimate contact area generated by the proof load application (yielded
surface in the contact area during proof load).
Max law
αint
Figure 24: Cyclic OPB stress results fct of interlink Figure 25: Material law influence on bending moment
angle curve
Direct comparison of calculated interlink contact stiffness derived from FEA to the one derived from the static test program
has also shown that FE results generally underestimate the interlink contact stiffness.
Sensitivity studies to several parameters were performed and it was determined that the material property was the most
influencing parameter.
For most FEA works performed, a Ramberg Osgood elastic-plastic material type of law had been used. Figure 25 shows that
using material laws based on the minimum or the maximum yield and ultimate tensile strength expected has a large influence
on the resulting OPB stiffness curve. At a later stage of the project some specific cyclic elastic plastic tests were performed on
small samples taken from chain links (tests performed by ENSAM Angers) to assess a more suitable hardening model for steel
grade R4 used for the fatigue tests.
For this new model, the plasticity is described by the Von Mises yield surface and associated plastic flow rule. The hardening
model is based on a combined non-linear isotropic and non-linear kinematic model.
The tests implemented were made using typical samples shown in Figure 26 that were taken from 84 mm grade R4 chain links.
A monotonic tensile test was first implemented up to sample failure to define the elastic material properties and the single
cycle plastic properties. Then imposed strain cyclic tests were performed up to sample failure for several imposed strain levels.
This enabled the definition of a stabilized dynamic (cyclical) stress-strain curve which is shown on Figure 27 together with the
monotonic static stress stain curve. The comparison of both curves highlights the significant softening that the material
undergoes as the cyclic stress strain curve is largely below the static stress strain curve. This is classical for such quenched and
tempered materials and tests results show a quick softening during the first 10 cycles and then a continuous softening until
failure.
OTC-25779-MS 11
Using this new material law in FEA has improved the resulting interlink contact stiffness to match the one derived from static
test program on chains. However, the FEA result remains non conservative.
As a summary from the various sensitivities studies that have been run during the various FEA scope of works:
• The following sensitivity studies on material properties were performed:
o Extreme (minimum & maximum) tensile properties were tested and were shown to have significant
influence on interlink stiffness
o The tensile properties homogeneity through the link section was assessed by both Vicinay and Ramnas
manufacturers by implementing tensile static tests on small samples taken at various locations within the
chain link section and no significant discrepancy was found
o The influence of the proof load application in the tensile properties definition was also been assessed by
comparing static tensile tests results from samples taken from proof loaded and non proof loaded chain links
o Finally, the influence of using a cyclic stress strain curve defined as described above as shown a significant
influence on FEA results as the hardening law contributes significantly in the stabilized state (residual stress,
shakedown effect).
• Sensitivity studies to real link shape were performed by comparing the nominal link shape to a “real fabricated pre-
proof load link shape” anticipating the plastic link deformation and shape change during proof load. This was done by
modelling the chain link in line with average measured dimensions before proof load during manufacturing process.
The real elliptical link shape in the crown area has also been taken into account as shown in Figure 28. The analysis
result has shown significant influence on the OPB Stress Concentration factors (SCFs) achieved in the crown area.
• The sensitivity to fabrication tolerances was studied and there was insignificant influence.
• The sensitivity to the interlink friction coefficient did not indicate significant influence on the OPB bending stiffness
for small interlink angle range. However, for larger interlink angle, the interlink friction coefficient directly
influences both rolling and sliding interlink locking modes.
• The sensitivity to experimental boundary conditions was also examined (as shown in Figure 21) and there was
insignificant influence
12 OTC-25779-MS
The final FEA scope of work consists of an FE model representing the complete 15 links sample configuration in the test
bench. The study was performed to simulate correct boundary conditions (as shown in Figure 29) and also to validate the
bending moment transfer function which will be used to derive the OPB bending moment at the interlink contact area (as
described previously in the static test post processing work) from the nominal OPB moment in the link straight part (where the
strain gauges were placed).
Using this test bench configuration, 10 cycles were run (up and down vertical displacement was applied on the center link as in
the static chain testing) in order to reach a stabilized cycle. The newly combined non-linear isotropic and non-linear kinematic
model was also used in the analyses to better capture the cyclic material behaviour.
One other important objective from the FEA work was to characterize the stress state at various locations on the link and to
define OPB SCFs at fatigue sensitive locations. The OPB SCF was defined as the ratio between the OPB stress at a given node
and the nominal OPB stress (derived from the OPB bending moment at this location and nominal bending inertia).
This objective was met with the various FEA studies. Initially, the maximum principal stress had been used as a criterion.
Analyses results have indicated that the maximum principal stresses due to OPB loading occur between locations B or B’ as
shown in Figure 32 (depending on link shape and loading conditions) rather at the beginning of the link curvature.
Additionally, studying variation of the principal stresses and shear stress during one full cycle on various locations has shown
that the stress state is multiaxial in the curved part of the link. Figure 30 and Figure 31 show that some significant shear stress
is concomitant with the maximum principal stress at location C and that the stress state at this location is not uniaxial.
Concurrently, the results of fatigue test program have indicated that all fatigue failures did start close to location C
corresponding well to the maximum shear stress location.
This is a strong indication that a multiaxial fatigue criterion would be more suitable for such OPB loading than a uniaxial
fatigue criterion based on the usual maximum principal stress.
This last FE model (test bench configuration described above) was also used to derive a complete stress tensor on the more
OTC-25779-MS 13
loaded OPB links. These stress tensors were used to run a specific multiaxiality study subcontracted to Andre Bignonnet
Consulting targeting to define a suitable multiaxial SCF.
A multiaxial SCF was proposed by Andre Bignonnet Consulting based on a recent multiaxial fatigue theoretical development
proposed by K Dang Van (see Ref [10] Ref [12] and Ref [11] ). The multiaxial fatigue theory is based on critical plane
theories to assess the macroscopic flaw initiation stage. In the Dang Van method, the assumption is made that the structure is
submitted to high cycles fatigue which implicitly means that the overall behaviour remains elastic. The main hypothesis is that
a fatigue crack will initiate if and only if the material doesn’t reach elastic shakedown locally. The elastic shakedown at the
mesoscopic scale (metallic grain size) may occur only if an elastically adapted mechanical state is reached at the scale of the
structure. The multiaxial fatigue criterion defined by K. Dang Van gives the critical loading path in a τ – p diagram (shear
stress as a function of hydrostatic pressure which is the hydrostatic part of the stress tensor p=I1/3=1/3(s1+s2+s3)).
The Dang Van criterion considers the maximum shear amplitude on the shear planes and selects the plane of the maximum
value. This quantity is associated to the concomitant hydrostatic pressure (invariant scalar).
.
{ }
max max [τ (n, cycle) + a * p (t )] ≤ b
t n
This criterion was applied on stress tensors derived from the last FE model and analyses (full 15 chain links in the test bench
configuration). The results have indicated that location C (shown in Figure 33) is always the most fatigue sensitive location for
all investigated cases.
The proposed design approach did consist in deriving an OPB stress range for a given bending moment in the contact area.
The proposed multiaxial SCF establish a relationship between a traditional uniaxial push pull fatigue S-N curve and a Dang
Van curve. With regard to the nominal uniaxial stress Sa,nom defined as constant amplitude push pull loading for a nominal
bending moment at each location, the following multiaxial SCF is defined:
14 OTC-25779-MS
(τ − α . p amp )
SCFmultiaxial =
a
Equation 6 : where τa is the shear amplitude, Pamp is the hydrostatic pressure amplitude
1 α
S a ,nom * −
2 3
and α is the Dang Van material slope which was defined by specific testing done on small samples machined from chain links
(as described in Ref [1] and Ref [2] .
This SCF accounts for the multiaxial specificity at the different locations considered. Employing this SCF, an equivalent
uniaxial stress range can be derived which implicitly accounts for the multiaxial stress state at given locations. This equivalent
uniaxial stress can then be added to the contribution from both potential tension and IPB uniaxial stress ranges in the same way
as it is done with uniaxial SCFs.
This multiaxial SCF is a dimensionless coefficient based on invariants allowing the design stress weighting. It remains
coherent to what is usually done for uniaxial analysis with TT and IPB cyclic loading. The multiaxial SCF accounts for added
effect of the stress tensor multiaxiality. It allows defining an equivalent uniaxial design stress, in the sense of the fatigue
damage, in the reference system (vertical axis of the reference S-N line).
Therefore for load case combination and cumulative damage purpose based on the uniaxial reference SN curve, one can derive
an equivalent uniaxial design stress at location i (C, B, B’) due to OPB:
The calculated SCF values have initially been calculated using the 15% MBL tension case as a reference system.
For other tension values, the reference system must be slightly adjusted to the new hydrostatic mean stress value, pmean at a
given location. The mean stress effect is small. However, it can be taken into account in the previous calculated SCF values by
multiplying it with a factor γTT as follow:
Note that γTT is the same for all locations as it is based on a correction from the reference mean hydrostatic stress.
The simplification of γTT parameter dependency to the OPB stress range is based on the calculated values for the most
representative load cases. This brings to one single value of 1.09 for a static tension of 25% MBL.
In conclusion, as results of various FEA studies implemented within the course of this project, it was highlighted that the FEA
was an effective tool to better understand the OPB locking mechanism and associated interlink contact stiffness. It provided a
method to derive suitable stress states at various locations and demonstrated that a multiaxial fatigue criterion was more
suitable than a criterion based on the (uniaxial) maximum principal stress. A multiaxiality study was specifically implemented
to propose a multiaxial SCF definition based on Dang Van fatigue criterion. Multiaxial OPB SCF values were consequently
proposed at various locations.
On another hand, the interlink contact stiffness derived from FEA was underestimated compared to the interlink contact
stiffness derived from the static test campaign. Therefore, it was recommended to use the interlink contact stiffness law
derived from tests results as it statistically accounts better for all uncertainties that can exist on real chain links.
Full scale chain fatigue test program description and main results
A full scale fatigue testing program was performed at SES’s laboratory using chain samples of 15 links. The primary objective
was to mimic the offshore fatigue failure and to obtain sufficient data to derive an OPB S-N curve. The sensitivity to the main
parameters (stress range, cyclic frequency, environment, chain size and mean static tension) was also assessed.
The fatigue testing program can be summarized as follow:
• 12 different fatigue tests were performed on full scale samples (2 in air and 10 in aerated seawater) without any
strong limitation on the number of cycles per sample (the cyclical limit was contractually defined on the total test
program, giving the JIP significant room for very long duration fatigue tests).
• One chain grade was tested: Grade R4
OTC-25779-MS 15
• 2 different chain diameters were tested (84 mm and 127 mm studless chains)
• The first samples were used to assess the environment effect (air / seawater) and frequency effect (0.2 Hz / 1 Hz)
• Four (4) different stress range levels were applied in testing the 84 mm chains and 3 different stress levels were
applied in testing the 127 mm chains.
The fatigue test bench previously shown on Figure 8 was specifically constructed to allow testing 2 chain samples
simultaneously, to test large chain and to withstand the significant fatigue loading of the fatigue testing program. Both
independent bays were equipped with a one Million LBS cylinder for applying the static tension and a vertical cylinder for
applying the vertical stroke (displacement) on the center link. The test set-up is the same as the one described for the static test
program. The OPB links L7 and L9 sustain the maximum OPB moment and therefore were expected to break first.
The table below summarizes the testing conditions for the 12 different fatigue tests completed.
Initially, the fatigue testing program was to be performed with three different mean static tension levels (25% MBL, 20%
MBL, 15% MBL) and with a maximum stress range targeted at 90% of the sliding threshold (i.e. sliding mode during the
fatigue tests is not desirable). In line with these requirements, the first two tests, samples F5 and F6 tested in air, were
performed with tension of 25% MBL. Fatigue failure occurred rapidly. The third test, sample F2, was performed with tension
of 15% MBL but failure still occurred too rapidly. It was then decided to keep the static tension value of 15% MBL for all
remaining fatigue tests and to adjust the stroke (vertical displacement) range to achieve targeted stress ranges. This decision
also removed the sensitivity to the mean stress effect as all tests were run with approximately the same mean stress level. The
mean stress effect was then accounted separately using the correction factor γTT described previously and, subsequently, it was
possible to derive the fatigue curves for various mean stress level.
The fatigue domain of interest includes the high cycles regime (low stress range and large number of cycles, typically above 2
millions). Due to schedule constrain it was necessary to perform the fatigue tests at a higher frequency (1 Hz) than typical
offshore conditions (0.1 to 0.2 Hz).
Moreover, it was initially proposed to perform all fatigue tests in air and then to derive an S-N curve in seawater using the
factors given by standard S-N curves. However, the validity of such extrapolation was challenged by previous JIP results, see
Ref [9] , in which it demonstrates that results of T-T fatigues tests on chains in air and in corrosive seawater show a different
trend (different slope). It was also demonstrated that tests performed in seawater at 0.2 Hz and 0.7 Hz indicated that testing
frequency significantly impact the test results; at 0.2 Hz testing frequency, the fatigue performance was lower than that of 0.7
Hz. . Further from Ref [9] when results from testing in air were extrapolated, they do not correlate well with results from
seawater testing. Lastly, the ratio of fatigue life between S-N curves in air and in seawater can vary from factors of 2 to 5
16 OTC-25779-MS
The targeted cyclic frequency was 1Hz apart for fatigue tests F1 which was intended to be run at a frequency of 0.2 Hz closer
to a real offshore frequency in order to assess the impact of this parameter. However, for large stroke range (F5 and F6) and
also for the largest chain size (F13, F14, F15), this cyclic frequency was not achievable (about 0.4 Hz was achieved).
The artificial seawater conditions are in line with ASTM definition (Seawater specific gravity range: 1.020 – 1.023, NaCl
concentration of 35g/L). In addition, air flow circulation was implemented to create aerated seawater conditions.
Two specific testing programs were conducted on small samples extracted from chain links with the objective of
independently assessing the influence of testing frequency increase in seawater environment on both crack initiation and crack
propagation stage.
See Ref [2] for further details on the tests that were performed on small samples in the chain OPB fatigue JIP. These tests
were intended to provide insight on the effect of increasing testing frequency to 1 Hz (instead of a normal offshore frequency
of 0.1Hz) for the full scale fatigue tests. This has been done by looking separately at the influence of the environment on crack
propagation rate and by assessing the fatigue endurance threshold for corroded and non corroded material (crack initiation
stage).
The program to assess the influence of environment on fatigue crack initiation was performed by CETIM. It consisted of
running rotative bending and torsional fatigue tests on typical samples (machined out of chain links, see Figure 36). Both non
corroded and corroded samples were tested to assess the influence of typical corrosion pits on crack initiation stage.
The program to assess the influence of environment on fatigue crack propagation was performed by TWI. It consisted of
frequency scanning tests in both seawater and in air environment on typical crack notched specimen (machined out of chain
links, see Figure 37).
A simple correction model was proposed from the work done by TWI for the crack propagation stage and the effect of the
corrosion status on crack initiation stage can be framed by fatigue endurance results comparison between corroded and non-
corroded samples. This provides an idea of the non-conservatism order of magnitude related to performing tests at higher
frequencies in a corrosive environment.
For each full scale fatigue test, the challenge was to assess which fatigue mechanism (crack initiation or crack propagation)
dominates. An appropriate correction must then be applied on the results, either on the number of cycles to failure or on the
stress range. The test results performed at 0.2Hz and at 1 Hz with similar stress range were also considered for deriving the
fore mentioned correction. After extensive analyses on the results of all full scale fatigue tests, the correction proposed was to
reduce the number of cycles to failure by 30% for all full scale tests performed in seawater at 1 Hz.
Going back to full scale fatigue testing, it is important to mention that all chain samples had been inspected by means of NDT
(links L7 and L9) to detect potential flaws. This was done in the chain manufacturing facility. All samples tested in seawater
were also first pre-soaked in a seawater bath for a minimum of two weeks.
Each sample was monitored with uniaxial strain gauges in the link straight part for the most loaded OPB links L7 and L9 and
with inclinometers on links L7, L8 and L9. The test was started with first running a minimum of 1000 cycles to remove the
rust in between chain links. Then a setting loop of 20 cycles reaching the interlink sliding was performed to determine the
target vertical stroke range (and the target stress range). On some samples, the strain gauges were kept throughout the duration
of fatigue test phase recording stress range data up to 1 million of cycles.
Among the12 fatigue tests run, 10 samples reached failure, i.e. at least one of the most loaded OPB link broke (completely
opened or nearly opened). The remaining two samples were tested for 10 million and 11 millions of cycles without reaching
OTC-25779-MS 17
failure. For each broken link, some fractographic assessment was conducted to determine the crack initiation location. The
failed cross section is first defined with an angular position in the link crown area measured from the start of the curvature
(alpha angle as shown in Figure 32 and Figure 38). Then, within the defined failed cross section, some closer analysis was
performed to assess the crack initiation point location that was determined with the use of beta angle as defined in Figure 33
and shown in Figure 39 and Figure 40).
The figure above shows a typical OPB chain link fatigue failure. The failure occurs in the bent area, approximately 50 degrees
inside the curvature (alpha angle) in the close proximity of the contact area border.
The following figures show typical failed cross sections. Most failed chain links have shown several competitive cracks that
propagated simultaneously (4 identical locations; from the top, from the bottom and on both link sides).
It was identified that the typical OPB crack start at approximately 50 degree from the top (or from the bottom) of the link, in
line with beta angle definition shown in Figure 33 (see also cracks shown in both Figure 39 and Figure 40).
Figure 39: Failed cross section example with one single Figure 40: Failed cross section example with two
crack symmetrical competitive cracks
18 OTC-25779-MS
Based on these fracture analysis, it was established that all failures (apart for both tests run in air with a very high stress range
where the crack initiation point has been associated with a flaw) started near the so called location C (as shown in Figure 32).
This location corresponds to the maximum shear stress location and is near the contact area border but at a short distance of it
(half a centimeter) which excludes fretting fatigue phenomenon. This location yields the worst result when multiaxial fatigue
criterion is applied (highest critical equivalent uniaxial stress leading to lower fatigue life). Therefore, this was a strong
indication that a multiaxial fatigue criterion predicted better the OPB fatigue behaviour of chain link and was more suitable to
use than a uniaxial fatigue criterion that would be based on the maximum principal stress.
The common practice for deterministic fatigue calculation in the offshore industry consists of summing individual fatigue
damages which is calculated with dedicated S-N curves (following the Palmgren Miner rule).
Ni
Equation 10 : FD = ∑ with
i N i , failure
as it would not be conservative to take no correction at all. The proposed value is deemed realistic based on the test results.
Note that this factor is applied on the number of cycles and can be likened to a safety factor.
Finally the stress range and number of cycles used to derive an S-N curve are summarized by both following equations.
0.15
d
Equation 13 : ∆σ OPB _ hot spot = ∆σ OPB _ M * Mbratio * SCFmultiaxial * d corr _ factor where d corr _ factor =
84
Equation 14 : N corrected = N test _ failure * Ftest _ freq _ correction
Based on the above definition, a dedicated OPB S-N curve has been derived from all seawater tests result accounting for all
primary failures, secondary failures and run-outs (run-outs with a stress range below the failure points were excluded). It
should be noted that the number of cycles associated to secondary failures and run-outs were conservatively not corrected even
if this does not correspond to a complete chain link opening. The stress averaging over the 4 gauges associated to the link
failure is deemed as a conservative step.
The data set selected is shown in the following figure (see Figure 41). It can be noticed that the data points correlate well with
the DNV B1 S-N curve for high stress ranges. However, for high cycle fatigue regime (low stress range and large number of
cycles) the test data points are significantly above the free corrosion curve prediction.
From this data set, a single slope S-N curve with a slope of 1/3 (m value equal to 3) was derived in the S-N curve diagram.
This slope has been forced to be in line with the slope proposed by most of the free corrosion fatigue curves for raw material.
This is also related to the common Paris law curve slope related to crack propagation stage.
A dual slope approach was also proposed based on the consideration of two data sets among the tests results:
• First part of the curve with a slope of 1/3 for high stress ranges (m1=3)
• Second part of the curve with a slope close to 1/5 based on the data points with large number of cycles (m2=4.94).
• The intercept between both curves was initially based on tests results but then the second part of the curve was
conservatively moved down so that the intercept occurs at 2 millions of cycles (on the design curve).
This slope change proposal was justified by tests results and it was also motivated by the fact that high cycle fatigue regime is
potentially driven by crack initiation mechanism that could justify a slope change. It was also demonstrated within the course
of this project that crack initiation fatigue curves of non-corroded and corroded samples (test program on small samples) have
different slopes.
It should be noted however that no consensus was reached within the course of this project on the dual slope proposal even if
benchmarking in-service units fatigue life with calculated OPB fatigue life using the dual slope provides better results than
using the single slope S-N curve.
The number of “high cycles fatigue data points” (typically above 1 or 2 millions of cycles) was indeed limited. The achieved
corroded condition of the chain links during seawater tests was also difficult to assess even if some significant corrosion pits
were found on chain links after cleaning for MPI inspection. These pits were globally in line with measured corrosion pits
from recovered chain links after years of service life (north sea environment).
It should be highlighted also that most of standard S-N curves are based on tests data made on samples tested up to only 1 or 2
millions of cycles maximum. It would therefore be valuable, albeit difficult and expensive, to implement further testing at
large number of cycles and in a corrosive environment.
20 OTC-25779-MS
Log ∆σOPB
Log Nfailure
Figure 41: S-N fatigue curve diagram
It can be noticed also that the standard deviation used to derive the design curve from the mean curve was conservatively
selected as the worst value between calculated standard deviation from test data set and standard value of 0.2 (standard
deviation on logN).
Finally, a reliability study implemented by Bureau Veritas (see Ref [4] ) has accounted for the reduced number of critical OPB
links (a few links only at each mooring line connection to the unit) compared to number of critical links for tension fatigue.
The output was to propose a minimum Fatigue Safety Factor of 3 (on the design life) for the single slope S-N curve with a
slope value of 1/3 (m equal to 3) and a minimum fatigue safety factor of 5 for a curve having a slope of 1/5 (m value of 5).
Chain OPB fatigue JIP main findings and design calculation parameters
Within the course of this JIP, an endeavour that took nearly seven years to complete, a greater understanding of chain OPB
fatigue mechanism has been achieved with emphasis of the following major points:
• A significant locking interlink contact stiffness was demonstrated as results of both full scale static and fatigue tests
programs
• The various interlink locking modes were identified
• An understanding of the various source of scatter during testing (locking point / mode,…) was achieved
• A sensitivity to the various design parameters (Tension, chain diameter, material properties, tolerances, etc.) was
successfully determined
• It has been demonstrated that the multiaxial stress state near the contact area justifies the crack initiation location
• Interlink contact stiffness law was defined using the tabulated parameters values provided by the least square method
or using the parametric normalized formulations:
OTC-25779-MS 21
Equation 19 : ∆σ OPB ,i = γ TT * SCFmultiaxial ,OPB ,i * ∆σ no min al ,i with SCFmultiaxial,OPBi, and γTT defined by Equation 6 and
Equation 9.
The following table summarized all SCFs proposed for OPB, Tension, and IPB loadings for studless chain links.
The OPB SCF was defined using a multiaxial fatigue criterion that accounts for the geometrical and contact OPB
characteristics. Recall that this OPB multiaxial SCF can be used to provide an equivalent uniaxial stress range as it is a
dimensionless coefficient based on invariants (such as for T-T and IPB SCF). As it applies a multiaxial fatigue criterion on a
uniaxial traction –compression curve, at the end it can be used like the other uniaxial SCF. The associated stress range can
consequently be directly added to other stress components (for tension and IPB loading).
This multiaxial SCF implicitly accounts for the multiaxial stress state at the given location. The hotspots ranking with this
definition is in line with the crack initiation starting at location C (near the contact area).
It is noted that a recently released Chain OPB fatigue Calculation Guidance from Bureau Veritas (see Ref [3] ) summarizes all
these design parameters.
22 OTC-25779-MS
Conclusion
This project succeeded in reproducing the OPB interlink contact stiffness and in demonstrating the associated fatigue
performance. This was the first full scale fatigue testing program that clearly demonstrated mooring chains can be OPB fatigue
sensitive when significant static tension are applied together with large mooring line motions. Through full scale testing under
real mooring line loads, it also identified the main parameters that play a role in the design of suitable mooring line
connections to offshore units.
Finally a breakthrough in the standard practice has been made by proposing a Stress Concentration Factors based on a
multiaxial fatigue criterion while keeping the standard deterministic S-N curve approach.
This project has also shown the need to perform more high cycle fatigue testing in a corrosive environment because there is a
strong indication that the global methodology conservatism can be related to over conservative contribution of the resistant
part of the fatigue calculation (through the fatigue curve performance for low stress range).
Acknowledgements
References
Ref [1] SBM Offshore -RD51254 – RRM86007A1 – Chain Out of Plane Bending JIP Report – L. Rampi, F. dewi – July
2013
Ref [2] Fatigue Design 2011 – Methodology to account for corrosive environment on accelerated fatigue test on mooring
chains within the Chain out of Plane Bending (OPB) Joint Industry Project (JIP) – L. Rampi, P. Vargas
Ref [3] Bureau Veritas - Guidance Note NI 604 DT R00E - Fatigue of Top Chain of Mooring Lines due to In-plane and
Out-of-plane Bending – October 2014
Ref [4] Bureau Veritas - RE\DTO\615\AGT – Reliability Analysis on Safety Factors for the Top Chain Combined Fatigue
- A. Gerthoffert
Ref [5] OTC2005-17238-PP – Failure of Chains by Bending on Deepwater Mooring Systems – P. Jean, K. Goessens, D.
L’Hostis
Ref [6] OMAE2005-67353-Out of Plane Bending Testing of Chain Links – C. Melis, P. Jean, P. Vargas
Ref [7] OMAE2005-67354 – FEA of Out of Plane Fatigue Mechanism of Chain Links – P. Vargas, P. Jean
Ref [8] OMAE2006-92488- Fatigue Testing of Out of Plane Bending Mechanism of Chain links – L. Rampi, P. Vargas
Ref [9] Fatigue of Mooring Chain in Air an Water – Results & Analysis – OTC8147-1996 – J. Stiff, D. Smith
Ref [10] K Dang Van, Macro-micro Approach in High cycle Multiaxial Fatigue, Advances in Multiaxial Fatigue, STP
1191, D.L. Mc Dowell and R. Ellis, Ed ASTM, 1993;
Ref [11] Ballard, Dang Van, Deperrois, Papadopoulos, 1995, High Cycle fatigue and Finite Element Analysis – Fatigue
and Fracture Engineering Material in Structures, Vol. 18, No 3
Ref [12] DANG VAN K., GRIVEAU B., MESSAGE O., 1989, “On a new multiaxial fatigue limit criterion: theory and
application”, Biaxial and Multiaxial Fatigue, EGF3, Mechanical Engineering Publications, pp 479-496.